aid stringlengths 9 15 | mid stringlengths 7 10 | abstract stringlengths 78 2.56k | related_work stringlengths 92 1.77k | ref_abstract dict |
|---|---|---|---|---|
1707.08668 | 2741718594 | Robots operating alongside humans in diverse, stochastic environments must be able to accurately interpret natural language commands. These instructions often fall into one of two categories: those that specify a goal condition or target state, and those that specify explicit actions, or how to perform a given task. Recent approaches have used reward functions as a semantic representation of goal-based commands, which allows for the use of a state-of-the-art planner to find a policy for the given task. However, these reward functions cannot be directly used to represent action-oriented commands. We introduce a new hybrid approach, the Deep Recurrent Action-Goal Grounding Network (DRAGGN), for task grounding and execution that handles natural language from either category as input, and generalizes to unseen environments. Our robot-simulation results demonstrate that a system successfully interpreting both goal-oriented and action-oriented task specifications brings us closer to robust natural language understanding for human-robot interaction. | There has been a broad and diverse set of work examining how best to interpret and execute natural language instructions on a robot platform . produce policies using language and expert trajectories based rewards, which allow for planning within a stochastic environment along with re-planning in case of failure. instead grounds language to trajectories satisfying the language specification. @cite_0 chose to ground language to constraints given to an external planner, which is a much smaller space to perform inference over than trajectories. formulate language grounding as a machine translation problem, treating propositional logic functions as both a machine language and reward function. Reward functions or cost functions can allow richer descriptions of trajectories than plain constraints, as they can describe preferential paths. Additionally, simplify the problem from one of machine translation to multi-class classification, learning a deep neural network to map arbitrary natural language instructions to the corresponding reward function. | {
"cite_N": [
"@cite_0"
],
"mid": [
"2099217059"
],
"abstract": [
"Robots that can be given instructions in spoken language need to be able to parse a natural language utterance quickly, determine its meaning, generate a goal representation from it, check whether the new goal conflicts with existing goals, and if acceptable, produce an action sequence to achieve the new goal (ideally being sensitive to the existing goals). In this paper, we describe an integrated robotic architecture that can achieve the above steps by translating natural language instructions incrementally and simultaneously into formal logical goal description and action languages, which can be used both to reason about the achievability of a goal as well as to generate new action scripts to pursue the goal. We demonstrate the implementation of our approach on a robot taking spoken natural language instructions in an office environment."
]
} |
1707.08721 | 2741133657 | In recent years, the performance of object detection has advanced significantly with the evolving deep convolutional neural networks. However, the state-of-the-art object detection methods still rely on accurate bounding box annotations that require extensive human labelling. Object detection without bounding box annotations, i.e, weakly supervised detection methods, are still lagging far behind. As weakly supervised detection only uses image level labels and does not require the ground truth of bounding box location and label of each object in an image, it is generally very difficult to distill knowledge of the actual appearances of objects. Inspired by curriculum learning, this paper proposes an easy-to-hard knowledge transfer scheme that incorporates easy web images to provide prior knowledge of object appearance as a good starting point. While exploiting large-scale free web imagery, we introduce a sophisticated labour free method to construct a web dataset with good diversity in object appearance. After that, semantic relevance and distribution relevance are introduced and utilized in the proposed curriculum training scheme. Our end-to-end learning with the constructed web data achieves remarkable improvement across most object classes especially for the classes that are often considered hard in other works. | : Traditional WSD methods like @cite_23 address this problem with multiple instance learning (MIL) @cite_24 , which treats each image as a bag and each proposal window in the image as an instance in the bag. A positive image contains at least one positive instance whereas a negative image contains only negative instances. Since MIL approaches alternate the processes between selecting a region of objects and using the selected region to learn the object appearance model, they are often sensitive to initialization and often get stuck in local optima. @cite_19 proposed a two-stream CNN structure named WSDDN to learn localization and recognition in dedicated streams respectively. These two streams share the common features from the earlier convolutional layers and one fully connected layer. It learns one detection stream to find the high responsive windows and one recognition stream to learn the appearance of the objects. In this way, the localization and recognition processes are decoupled. Similarly, @cite_20 also uses the two stream structure and involves the context feature in the localization stream. | {
"cite_N": [
"@cite_24",
"@cite_19",
"@cite_20",
"@cite_23"
],
"mid": [
"2110119381",
"2101611867",
"2519284461",
"2133324800"
],
"abstract": [
"The multiple instance problem arises in tasks where the training examples are ambiguous: a single example object may have many alternative feature vectors (instances) that describe it, and yet only one of those feature vectors may be responsible for the observed classification of the object. This paper describes and compares three kinds of algorithms that learn axis-parallel rectangles to solve the multiple instance problem. Algorithms that ignore the multiple instance problem perform very poorly. An algorithm that directly confronts the multiple instance problem (by attempting to identify which feature vectors are responsible for the observed classifications) performs best, giving 89 correct predictions on a musk odor prediction task. The paper also illustrates the use of artificial data to debug and compare these algorithms.",
"Weakly supervised learning of object detection is an important problem in image understanding that still does not have a satisfactory solution. In this paper, we address this problem by exploiting the power of deep convolutional neural networks pre-trained on large-scale image-level classification tasks. We propose a weakly supervised deep detection architecture that modifies one such network to operate at the level of image regions, performing simultaneously region selection and classification. Trained as an image classifier, the architecture implicitly learns object detectors that are better than alternative weakly supervised detection systems on the PASCAL VOC data. The model, which is a simple and elegant end-to-end architecture, outperforms standard data augmentation and fine-tuning techniques for the task of image-level classification as well.",
"We aim to localize objects in images using image-level supervision only. Previous approaches to this problem mainly focus on discriminative object regions and often fail to locate precise object boundaries. We address this problem by introducing two types of context-aware guidance models, additive and contrastive models, that leverage their surrounding context regions to improve localization. The additive model encourages the predicted object region to be supported by its surrounding context region. The contrastive model encourages the predicted object region to be outstanding from its surrounding context region. Our approach benefits from the recent success of convolutional neural networks for object recognition and extends Fast R-CNN to weakly supervised object localization. Extensive experimental evaluation on the PASCAL VOC 2007 and 2012 benchmarks shows that our context-aware approach significantly improves weakly supervised localization and detection.",
"Object category localization is a challenging problem in computer vision. Standard supervised training requires bounding box annotations of object instances. This time-consuming annotation process is sidestepped in weakly supervised learning. In this case, the supervised information is restricted to binary labels that indicate the absence presence of object instances in the image, without their locations. We follow a multiple-instance learning approach that iteratively trains the detector and infers the object locations in the positive training images. Our main contribution is a multi-fold multiple instance learning procedure, which prevents training from prematurely locking onto erroneous object locations. This procedure is particularly important when using high-dimensional representations, such as Fisher vectors and convolutional neural network features. We also propose a window refinement method, which improves the localization accuracy by incorporating an objectness prior. We present a detailed experimental evaluation using the PASCAL VOC 2007 dataset, which verifies the effectiveness of our approach."
]
} |
1707.08721 | 2741133657 | In recent years, the performance of object detection has advanced significantly with the evolving deep convolutional neural networks. However, the state-of-the-art object detection methods still rely on accurate bounding box annotations that require extensive human labelling. Object detection without bounding box annotations, i.e, weakly supervised detection methods, are still lagging far behind. As weakly supervised detection only uses image level labels and does not require the ground truth of bounding box location and label of each object in an image, it is generally very difficult to distill knowledge of the actual appearances of objects. Inspired by curriculum learning, this paper proposes an easy-to-hard knowledge transfer scheme that incorporates easy web images to provide prior knowledge of object appearance as a good starting point. While exploiting large-scale free web imagery, we introduce a sophisticated labour free method to construct a web dataset with good diversity in object appearance. After that, semantic relevance and distribution relevance are introduced and utilized in the proposed curriculum training scheme. Our end-to-end learning with the constructed web data achieves remarkable improvement across most object classes especially for the classes that are often considered hard in other works. | In this research, we use WSDDN @cite_19 as an example to evaluate our learning method. Since WSDDN separates recognition and localization into two individual streams, it introduces additional degrees of freedom while optimizing the model, and hence it is hard to train at the early stage. It is also sensitive to initialization. Thus, in this work we propose to explicitly provide good initialization during the training process in an easy-to-hard manner. Note that although we utilize WSDDN as our baseline, our learning scheme is general and can be applied to other WSD methods as well. | {
"cite_N": [
"@cite_19"
],
"mid": [
"2101611867"
],
"abstract": [
"Weakly supervised learning of object detection is an important problem in image understanding that still does not have a satisfactory solution. In this paper, we address this problem by exploiting the power of deep convolutional neural networks pre-trained on large-scale image-level classification tasks. We propose a weakly supervised deep detection architecture that modifies one such network to operate at the level of image regions, performing simultaneously region selection and classification. Trained as an image classifier, the architecture implicitly learns object detectors that are better than alternative weakly supervised detection systems on the PASCAL VOC data. The model, which is a simple and elegant end-to-end architecture, outperforms standard data augmentation and fine-tuning techniques for the task of image-level classification as well."
]
} |
1707.08721 | 2741133657 | In recent years, the performance of object detection has advanced significantly with the evolving deep convolutional neural networks. However, the state-of-the-art object detection methods still rely on accurate bounding box annotations that require extensive human labelling. Object detection without bounding box annotations, i.e, weakly supervised detection methods, are still lagging far behind. As weakly supervised detection only uses image level labels and does not require the ground truth of bounding box location and label of each object in an image, it is generally very difficult to distill knowledge of the actual appearances of objects. Inspired by curriculum learning, this paper proposes an easy-to-hard knowledge transfer scheme that incorporates easy web images to provide prior knowledge of object appearance as a good starting point. While exploiting large-scale free web imagery, we introduce a sophisticated labour free method to construct a web dataset with good diversity in object appearance. After that, semantic relevance and distribution relevance are introduced and utilized in the proposed curriculum training scheme. Our end-to-end learning with the constructed web data achieves remarkable improvement across most object classes especially for the classes that are often considered hard in other works. | : This paper is also related to those works on learning from weak or noisy labels @cite_22 @cite_12 @cite_21 @cite_1 . In @cite_22 , they proposed a classifier-based cleaning process to deal with the noisy labels. They first train a classification model on images with higher confidence and then use this model to filter the outliers in the rest of images. Later, with incorporation of CNN, novel loss layer is introduced to the deep network in @cite_12 . In @cite_21 , web images are separated into easy images (Google) and hard images (Flickr). They build a knowledge graph on easy web images and use the graph as a semantic constraint to deal with the possible label-flip noises during training of harder web images. Similarly, @cite_1 learns the mutual relationship to suppress the feedback of noises during the back propagation. These works emphasize their methods to lessen the impact by outliers during the training process. In our work, apart from the outliers, we also consider distribution mismatch problem since we acquire web data that are from completely different information source with discrepant distribution compared to target dataset. | {
"cite_N": [
"@cite_21",
"@cite_1",
"@cite_22",
"@cite_12"
],
"mid": [
"2950931866",
"2213472736",
"2081613070",
"2121056381"
],
"abstract": [
"We present an approach to utilize large amounts of web data for learning CNNs. Specifically inspired by curriculum learning, we present a two-step approach for CNN training. First, we use easy images to train an initial visual representation. We then use this initial CNN and adapt it to harder, more realistic images by leveraging the structure of data and categories. We demonstrate that our two-stage CNN outperforms a fine-tuned CNN trained on ImageNet on Pascal VOC 2012. We also demonstrate the strength of webly supervised learning by localizing objects in web images and training a R-CNN style detector. It achieves the best performance on VOC 2007 where no VOC training data is used. Finally, we show our approach is quite robust to noise and performs comparably even when we use image search results from March 2013 (pre-CNN image search era).",
"The development of deep learning has empowered machines with comparable capability of recognizing limited image categories to human beings. However, most existing approaches heavily rely on human-curated training data, which hinders the scalability to large and unlabeled vocabularies in image tagging. In this paper, we propose a weakly-supervised deep learning model which can be trained from the readily available Web images to relax the dependence on human labors and scale up to arbitrary tags (categories). Specifically, based on the assumption that features of true samples in a category tend to be similar and noises tend to be variant, we embed the feature map of the last deep layer into a new affinity representation, and further minimize the discrepancy between the affinity representation and its low-rank approximation. The discrepancy is finally transformed into the objective function to give relevance feedback to back propagation. Experiments show that we can achieve a performance gain of 14.0 in terms of a semantic-based relevance metric in image tagging with 63,043 tags from the WordNet, against the typical deep model trained on the ImageNet 1,000 vocabulary set.",
"Recognition is graduating from labs to real-world applications. While it is encouraging to see its potential being tapped, it brings forth a fundamental challenge to the vision researcher: scalability. How can we learn a model for any concept that exhaustively covers all its appearance variations, while requiring minimal or no human supervision for compiling the vocabulary of visual variance, gathering the training images and annotations, and learning the models? In this paper, we introduce a fully-automated approach for learning extensive models for a wide range of variations (e.g. actions, interactions, attributes and beyond) within any concept. Our approach leverages vast resources of online books to discover the vocabulary of variance, and intertwines the data collection and modeling steps to alleviate the need for explicit human supervision in training the models. Our approach organizes the visual knowledge about a concept in a convenient and useful way, enabling a variety of applications across vision and NLP. Our online system has been queried by users to learn models for several interesting concepts including breakfast, Gandhi, beautiful, etc. To date, our system has models available for over 50, 000 variations within 150 concepts, and has annotated more than 10 million images with bounding boxes.",
"Current state-of-the-art deep learning systems for visual object recognition and detection use purely supervised training with regularization such as dropout to avoid overfitting. The performance depends critically on the amount of labeled examples, and in current practice the labels are assumed to be unambiguous and accurate. However, this assumption often does not hold; e.g. in recognition, class labels may be missing; in detection, objects in the image may not be localized; and in general, the labeling may be subjective. In this work we propose a generic way to handle noisy and incomplete labeling by augmenting the prediction objective with a notion of consistency. We consider a prediction consistent if the same prediction is made given similar percepts, where the notion of similarity is between deep network features computed from the input data. In experiments we demonstrate that our approach yields substantial robustness to label noise on several datasets. On MNIST handwritten digits, we show that our model is robust to label corruption. On the Toronto Face Database, we show that our model handles well the case of subjective labels in emotion recognition, achieving state-of-the- art results, and can also benefit from unlabeled face images with no modification to our method. On the ILSVRC2014 detection challenge data, we show that our approach extends to very deep networks, high resolution images and structured outputs, and results in improved scalable detection."
]
} |
1707.08148 | 2776682023 | Current image transformation and recoloring algorithms try to introduce artistic effects in the photographed images, based on user input of target image(s) or selection of pre-designed filters. These manipulations, although intended to enhance the impact of an image on the viewer, do not include the option of image transformation by specifying the affect information. In this paper we present an automatic image-transformation method that transforms the source image such that it can induce an emotional affect on the viewer, as desired by the user. Our proposed novel image emotion transfer algorithm does not require a user-specified target image. The proposed algorithm uses features extracted from top layers of deep convolutional neural network and the user-specified emotion distribution to select multiple target images from an image database for color transformation, such that the resultant image has desired emotional impact. Our method can handle more diverse set of photographs than the previous methods. We conducted a detailed user study showing the effectiveness of our proposed method. A discussion and reasoning of failure cases has also been provided, indicating inherent limitation of color-transfer based methods in the use of emotion assignment. | Image color transformation has been employed for variety of tasks, from colorization of the gray scale images to color de-rendering and image transfer. Where colorization of gray-scale images @cite_10 and color de-rendering @cite_9 use learned models for the task (requiring only the input image); most of the image transfer algorithms require both source and target images as input. Pouli and Reinhard @cite_19 use the histogram matching algorithm to transfer the color and tone from the target image to the source image. @cite_8 recolorizes the image such that style of the target image is transferred to the input image. Recently used deep convolutional network trained for object detection to transfer style of the target image to the source image. They claim that their algorithm understand how to separate the content and style because they use appropriate combination of high level and low-level features. However, their transformation is style transfer at the cost of realism. | {
"cite_N": [
"@cite_19",
"@cite_9",
"@cite_10",
"@cite_8"
],
"mid": [
"2065740805",
"2049380881",
"",
"2042465851"
],
"abstract": [
"Image manipulation takes many forms. A powerful approach involves image adjustment by example. To make color edits more intuitive, the intelligent transfer of a user-specified target image's color palette can achieve a multitude of creative effects, provided the user is supplied with a small set of straightforward parameters. We present a novel histogram reshaping technique which allows significantly more control than previous methods. Given that the user is free to chose any image as the target, the process of steering the algorithm becomes artistic. Moreover, we show for the first time that creative tone reproduction can be achieved by matching a high dynamic range image against a low dynamic range target.",
"Consumer-level digital cameras typically post-process raw captured image data to produce enhanced visually appealing output RGB images. Post-processing operations include color gamut compression, tone mapping and other non-linear color corrections. However, raw image data is needed for many computer vision applications such as photometric stereo, shape from shading, and color constancy. Recovering raw image data from RGB images is complicated by the high non-linearity of the post-processing operations. In this paper, we propose a coupled dictionary scheme to model the relationship between the raw and RGB color image spaces of consumer cameras. Dictionary learning is regularized by sparsity constraints on feature representation. As well, we explore a more elaborate variant of coupled dictionary schemes that models the feature coupling more accurately. We test the proposed dictionary learning schemes on many commercial camera datasets. Our experimental results show accurate recovery of raw image data that looks visually indistinguishable from the ground truth.",
"",
"In this paper, a color transfer framework to evoke different emotions for images based on color combinations is proposed. The purpose of this color transfer is to change the “look and feel” of images, i.e., evoking different emotions. Colors are confirmed as the most attractive factor in images. In addition, various studies in both art and science areas have concluded that other than single color, color combinations are necessary to evoke specific emotions. Therefore, we propose a novel framework to transfer color of images based on color combinations, using a predefined color emotion model. The contribution of this new framework is threefold. First, users do not need to provide reference images as used in traditional color transfer algorithms. In most situations, users may not have enough esthetic knowledge or path to choose desired reference images. Second, because of the usage of color combinations instead of single color for emotions, a new color transfer algorithm that does not require an image library is proposed. Third, again because of the usage of color combinations, artifacts that are normally seen in traditional frameworks using single color are avoided. We present encouraging results generated from this new framework and its potential in several possible applications including color transfer of photos and paintings."
]
} |
1707.08148 | 2776682023 | Current image transformation and recoloring algorithms try to introduce artistic effects in the photographed images, based on user input of target image(s) or selection of pre-designed filters. These manipulations, although intended to enhance the impact of an image on the viewer, do not include the option of image transformation by specifying the affect information. In this paper we present an automatic image-transformation method that transforms the source image such that it can induce an emotional affect on the viewer, as desired by the user. Our proposed novel image emotion transfer algorithm does not require a user-specified target image. The proposed algorithm uses features extracted from top layers of deep convolutional neural network and the user-specified emotion distribution to select multiple target images from an image database for color transformation, such that the resultant image has desired emotional impact. Our method can handle more diverse set of photographs than the previous methods. We conducted a detailed user study showing the effectiveness of our proposed method. A discussion and reasoning of failure cases has also been provided, indicating inherent limitation of color-transfer based methods in the use of emotion assignment. | @cite_4 instead of over-segmenting the input image use semantic segmentation. For each input-image-segment, semantically compatible segments are searched from the database while minimizing on position, scale, lightness of the segment and closeness of user supplied Valance Arousal (VA) score to VA score of the image to whom selected segment belongs to. Due to their reliance on the semantic segmentation they constraint themselves to the landscape images, they also rely on VA scores which are not easily interpret-able by the humans and each segment's color is changed by just changing the mean value of the segment. | {
"cite_N": [
"@cite_4"
],
"mid": [
"2544282078"
],
"abstract": [
"We introduce an affective image recoloring method for changing the overall mood in the image in a numerically measurable way. Given a semantically segmented source image and a target emotion, our system finds reference image segments from the collection of images that have been tagged via crowdsourcing with numerically measured emotion labels. We then recolorize the source segments using colors from the selected target segments while preserving the gradient of the source image to generate a seamless and natural result. User study confirms the effectiveness of our method in accomplishing the stated goal of altering the mood of the image to match the target emotion level."
]
} |
1707.08148 | 2776682023 | Current image transformation and recoloring algorithms try to introduce artistic effects in the photographed images, based on user input of target image(s) or selection of pre-designed filters. These manipulations, although intended to enhance the impact of an image on the viewer, do not include the option of image transformation by specifying the affect information. In this paper we present an automatic image-transformation method that transforms the source image such that it can induce an emotional affect on the viewer, as desired by the user. Our proposed novel image emotion transfer algorithm does not require a user-specified target image. The proposed algorithm uses features extracted from top layers of deep convolutional neural network and the user-specified emotion distribution to select multiple target images from an image database for color transformation, such that the resultant image has desired emotional impact. Our method can handle more diverse set of photographs than the previous methods. We conducted a detailed user study showing the effectiveness of our proposed method. A discussion and reasoning of failure cases has also been provided, indicating inherent limitation of color-transfer based methods in the use of emotion assignment. | For our experimental settings, we use the dataset introduced by @cite_2 where discrete emotion distribution is associated with each image. We use features extracted from the top layer of Deep CNN that capture both the content and spatial context information including the color and texture information at different locations of the image. We are not constrained by the semantic segmentation of an image and allow much more control of desired emotion by providing user ability to pick any discrete distribution of the emotions. Emotion distribution remains human interpret-able with the freedom to choose variety of combination of emotions. | {
"cite_N": [
"@cite_2"
],
"mid": [
"1950412479"
],
"abstract": [
"This paper explores two new aspects of photos and human emotions. First, we show through psychovisual studies that different people have different emotional reactions to the same image, which is a strong and novel departure from previous work that only records and predicts a single dominant emotion for each image. Our studies also show that the same person may have multiple emotional reactions to one image. Predicting emotions in “distributions” instead of a single dominant emotion is important for many applications. Second, we show not only that we can often change the evoked emotion of an image by adjusting color tone and texture related features but also that we can choose in which “emotional direction” this change occurs by selecting a target image. In addition, we present a new database, Emotion6, containing distributions of emotions."
]
} |
1707.08148 | 2776682023 | Current image transformation and recoloring algorithms try to introduce artistic effects in the photographed images, based on user input of target image(s) or selection of pre-designed filters. These manipulations, although intended to enhance the impact of an image on the viewer, do not include the option of image transformation by specifying the affect information. In this paper we present an automatic image-transformation method that transforms the source image such that it can induce an emotional affect on the viewer, as desired by the user. Our proposed novel image emotion transfer algorithm does not require a user-specified target image. The proposed algorithm uses features extracted from top layers of deep convolutional neural network and the user-specified emotion distribution to select multiple target images from an image database for color transformation, such that the resultant image has desired emotional impact. Our method can handle more diverse set of photographs than the previous methods. We conducted a detailed user study showing the effectiveness of our proposed method. A discussion and reasoning of failure cases has also been provided, indicating inherent limitation of color-transfer based methods in the use of emotion assignment. | red shows transformed images generated by providing both source and target images as in @cite_2 . As clear from the results, method works satisfactorily for certain pair of images. For example transformed image in the first column of has captured the energy and joy represented in the target image. However, as shown in last two columns, the target images have not changed their emotional class, although the intensity of their dominant emotion class has been reduced. One explanation for the failure in these cases is the mismatch between the HLCs in the source and target images, and it is clear that when there is a serious mismatch between the HLCs, it is difficult to transfer the emotional content of one image to the other by simply manipulating low-level features such as the color. In this paper, we propose a practical solution to the image emotion transfer using the proposed affective analysis. | {
"cite_N": [
"@cite_2"
],
"mid": [
"1950412479"
],
"abstract": [
"This paper explores two new aspects of photos and human emotions. First, we show through psychovisual studies that different people have different emotional reactions to the same image, which is a strong and novel departure from previous work that only records and predicts a single dominant emotion for each image. Our studies also show that the same person may have multiple emotional reactions to one image. Predicting emotions in “distributions” instead of a single dominant emotion is important for many applications. Second, we show not only that we can often change the evoked emotion of an image by adjusting color tone and texture related features but also that we can choose in which “emotional direction” this change occurs by selecting a target image. In addition, we present a new database, Emotion6, containing distributions of emotions."
]
} |
1707.08147 | 2740873362 | This paper addresses the problem of mixed initiative, shared control for master-slave grasping and manipulation. We propose a novel system, in which an autonomous agent assists a human in teleoperating a remote slave arm gripper, using a haptic master device. Our system is designed to exploit the human operator's expertise in selecting stable grasps (still an open research topic in autonomous robotics). Meanwhile, a-priori knowledge of: i) the slave robot kinematics, and ii) the desired post-grasp manipulative trajectory, are fed to an autonomous agent which transmits force cues to the human, to encourage maximally manipulable grasp pose selections. Specifically, the autonomous agent provides force cues to the human, during the reach-to-grasp phase, which encourage the human to select grasp poses which maximise manipulation capability during the post-grasp object manipulation phase. We introduce a task-relevant velocity manipulability cost function (TOV), which is used to identify the maximum kinematic capability of a manipulator during post-grasp motions, and feed this back as force cues to the human during the pre-grasp phase. We show that grasps which minimise TOV result in significantly reduced control effort of the manipulator, compared to other feasible grasps. We demonstrate the effectiveness of our approach by experiments with both real and simulated robots. | Grasping and manipulative motion planning have been widely studied in the robotics literature @cite_22 @cite_5 . Most of the studies, however, focused either on the first phase of grasping and manipulation, namely approach phase @cite_28 , or on the second phase, namely autonomously manoeuvring the manipulator @cite_6 . | {
"cite_N": [
"@cite_28",
"@cite_5",
"@cite_22",
"@cite_6"
],
"mid": [
"1999156278",
"2293467699",
"2563462039",
""
],
"abstract": [
"We consider the problem of detecting robotic grasps in an RGB-D view of a scene containing objects. In this work, we apply a deep learning approach to solve this problem, which avoids time-consuming hand-design of features. This presents two main challenges. First, we need to evaluate a huge number of candidate grasps. In order to make detection fast and robust, we present a two-step cascaded system with two deep networks, where the top detections from the first are re-evaluated by the second. The first network has fewer features, is faster to run, and can effectively prune out unlikely candidate grasps. The second, with more features, is slower but has to run only on the top few detections. Second, we need to handle multimodal inputs effectively, for which we present a method that applies structured regularization on the weights based on multimodal group regularization. We show that our method improves performance on an RGBD robotic grasping dataset, and can be used to successfully execute grasps on two different robotic platforms.",
"We describe a learning-based approach to hand-eye coordination for robotic grasping from monocular images. To learn hand-eye coordination for grasping, we trained a large convolutional neural network to predict the probability that task-space motion of the gripper will result in successful grasps, using only monocular camera images and independently of camera calibration or the current robot pose. This requires the network to observe the spatial relationship between the gripper and objects in the scene, thus learning hand-eye coordination. We then use this network to servo the gripper in real time to achieve successful grasps. To train our network, we collected over 800,000 grasp attempts over the course of two months, using between 6 and 14 robotic manipulators at any given time, with differences in camera placement and hardware. Our experimental evaluation demonstrates that our method achieves effective real-time control, can successfully grasp novel objects, and corrects mistakes by continuous servoing.",
"In this paper, we present a novel framework to drive automatic robotic grasp by matching camera captured RGB-D data with 3D meshes, on which prior knowledge for grasp is pre-defined for each object type. The proposed framework consists of two modules, namely, pre-defining grasping knowledge for each type of object shape on 3D meshes, and automatic robotic grasping by matching RGB-D data with pre-defined 3D meshes. In the first module, we scan 3D meshes for typical object shapes and pre-define grasping regions for each 3D shape surface, which will be considered as the prior knowledge for guiding automatic robotic grasp. In the second module, for each RGB-D image captured by a depth camera, we recognize 2D shape of the object in it by an SVM classifier, and then segment it from background using depth data. Next, we propose a new algorithm to match the segmented RGB-D shape with predefined 3D meshes to guide robotic self-location and grasp by an automatic way. Our experimental results show that the proposed framework is particularly useful to guide camera based robotic grasp.",
""
]
} |
1707.08147 | 2740873362 | This paper addresses the problem of mixed initiative, shared control for master-slave grasping and manipulation. We propose a novel system, in which an autonomous agent assists a human in teleoperating a remote slave arm gripper, using a haptic master device. Our system is designed to exploit the human operator's expertise in selecting stable grasps (still an open research topic in autonomous robotics). Meanwhile, a-priori knowledge of: i) the slave robot kinematics, and ii) the desired post-grasp manipulative trajectory, are fed to an autonomous agent which transmits force cues to the human, to encourage maximally manipulable grasp pose selections. Specifically, the autonomous agent provides force cues to the human, during the reach-to-grasp phase, which encourage the human to select grasp poses which maximise manipulation capability during the post-grasp object manipulation phase. We introduce a task-relevant velocity manipulability cost function (TOV), which is used to identify the maximum kinematic capability of a manipulator during post-grasp motions, and feed this back as force cues to the human during the pre-grasp phase. We show that grasps which minimise TOV result in significantly reduced control effort of the manipulator, compared to other feasible grasps. We demonstrate the effectiveness of our approach by experiments with both real and simulated robots. | There are a few numbers of studies on jointly considering the problem of grasping an object, manipulating it and delivering it to the desired pose. For example, two-phase optimisations were used in @cite_8 to generate the contact necessary for making a stable grasp on an object and to find the optimal object path that can be followed, given the optimal grasping configuration. In contrast, @cite_14 @cite_3 studied the optimal grasps resulting in a maximum manipulability at initial grasp configuration. Similarly, @cite_10 @cite_9 showed that different grasps can result in different task-oriented force manipulabilities as well as different torque efforts over the post-grasp motions. However, the main assumption of these works is that a planner can generate many stable grasp poses. Nonetheless, a reliable autonomous system has not yet developed that generates a stable grasp for an arbitrary object in a real world example. This is still an open research topic in autonomous robotics @cite_22 . | {
"cite_N": [
"@cite_14",
"@cite_22",
"@cite_8",
"@cite_9",
"@cite_3",
"@cite_10"
],
"mid": [
"2054329580",
"2563462039",
"1987985958",
"2300618187",
"",
""
],
"abstract": [
"The ability to grasp large objects with both hands enables bimanual robot systems to fully employ their capabilities in human-centered environments. Hence, algorithms are needed to precompute bimanual grasping configurations that can be used online to efficiently create whole body grasps. In this work we present a bimanual grasp planner that can be used to build a set of grasps together with manipulability information for a given object. For efficient grasp planning precomputed reachability information and a beneficial object representation, based on medial axis descriptions, are used. Since bimanual grasps may suffer from low manipulability, caused by a closed kinematic chain, we show how the manipulability of a bimanual grasp can be used as a quality measure. Therefore, manipulability clusters are introduced as an efficient way to approximatively describe the manipulability of a given bimanual grasp. The proposed approach is evaluated with a reference implementation, based on Simox [1], for the humanoid robot ARMAR-III [2]. Since the presented algorithms are robot-independent, there are no limitations for using this planner on other robot systems.",
"In this paper, we present a novel framework to drive automatic robotic grasp by matching camera captured RGB-D data with 3D meshes, on which prior knowledge for grasp is pre-defined for each object type. The proposed framework consists of two modules, namely, pre-defining grasping knowledge for each type of object shape on 3D meshes, and automatic robotic grasping by matching RGB-D data with pre-defined 3D meshes. In the first module, we scan 3D meshes for typical object shapes and pre-define grasping regions for each 3D shape surface, which will be considered as the prior knowledge for guiding automatic robotic grasp. In the second module, for each RGB-D image captured by a depth camera, we recognize 2D shape of the object in it by an SVM classifier, and then segment it from background using depth data. Next, we propose a new algorithm to match the segmented RGB-D shape with predefined 3D meshes to guide robotic self-location and grasp by an automatic way. Our experimental results show that the proposed framework is particularly useful to guide camera based robotic grasp.",
"Many manipulation planning problems involve several related sub-problems, such as the selection of grasping points on an object, choice of hand posture, and determination of the arm's configuration and evolving trajectory. Traditionally, these planning sub-problems have been handled separately, potentially leading to sub-optimal, or even infeasible, combinations of the individually determined solutions. This paper formulates the combined problem of grasp contact selection, grasp force optimization, and manipulator arm hand trajectory planning as a problem in optimal control. That is, the locally optimal trajectory for the manipulator, hand mechanism, and contact locations are determined during the pre-grasping, grasping, and subsequent object transport phase. Additionally, a barrier function approach allows for non-feasible grasps to be optimized, enlarging the region of convergence for the algorithm. A simulation of a simple planar object manipulation task is used to illustrate and validate the approach.",
"This paper presents a method for one-shot learning of dexterous grasps and grasp generation for novel objects. A model of each grasp type is learned from a single kinesthetic demonstration and several types are taught. These models are used to select and generate grasps for unfamiliar objects. Both the learning and generation stages use an incomplete point cloud from a depth camera, so no prior model of an object shape is used. The learned model is a product of experts, in which experts are of two types. The first type is a contact model and is a density over the pose of a single hand link relative to the local object surface. The second type is the hand-configuration model and is a density over the whole-hand configuration. Grasp generation for an unfamiliar object optimizes the product of these two model types, generating thousands of grasp candidates in under 30 seconds. The method is robust to incomplete data at both training and testing stages. When several grasp types are considered the method selects the highest-likelihood grasp across all the types. In an experiment, the training set consisted of five different grasps and the test set of 45 previously unseen objects. The success rate of the first-choice grasp is 84.4 or 77.7 if seven views or a single view of the test object are taken, respectively.",
"",
""
]
} |
1707.08350 | 2740930075 | We present a novel deep learning framework that models the scene dependent image processing inside cameras. Often called as the radiometric calibration, the process of recovering RAW images from processed images (JPEG format in the sRGB color space) is essential for many computer vision tasks that rely on physically accurate radiance values. All previous works rely on the deterministic imaging model where the color transformation stays the same regardless of the scene and thus they can only be applied for images taken under the manual mode. In this paper, we propose a data-driven approach to learn the scene dependent and locally varying image processing inside cameras under the automode. Our method incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers. The results show that we can model the imaging pipeline of different cameras that operate under the automode accurately in both directions (from RAW to sRGB, from sRGB to RAW) and we show how we can apply our method to improve the performance of image deblurring. | In the early works of radiometric calibration, the relationship between the scene radiance and the image value was explained just by a tonemapping function called the camera response function. Different models of the response function @cite_10 @cite_19 @cite_26 as well as robust estimation algorithms @cite_27 @cite_25 @cite_29 were introduced. More comprehensive reviews of the radiometric calibration literature is presented in @cite_8 . | {
"cite_N": [
"@cite_26",
"@cite_8",
"@cite_29",
"@cite_19",
"@cite_27",
"@cite_10",
"@cite_25"
],
"mid": [
"2130700878",
"2127701282",
"2133934946",
"1867965868",
"2069281566",
"2104745266",
"2123315723"
],
"abstract": [
"A simple algorithm is described that computes the radiometric response function of an imaging system, from images of an arbitrary scene taken using different exposures. The exposure is varied by changing either the aperture setting or the shutter speed. The algorithm does not require precise estimates of the exposures used. Rough estimates of the ratios of the exposures (e.g. F-number settings on an inexpensive lens) are sufficient for accurate recovery of the response function as well as the actual exposure ratios. The computed response function is used to fuse the multiple images into a single high dynamic range radiance image. Robustness is tested using a variety of scenes and cameras as well as noisy synthetic images generated using 100 randomly selected response curves. Automatic rejection of image areas that have large vignetting effects or temporal scene variations make the algorithm applicable to not just photographic but also video cameras.",
"We present a study of in-camera image processing through an extensive analysis of more than 10,000 images from over 30 cameras. The goal of this work is to investigate if image values can be transformed to physically meaningful values, and if so, when and how this can be done. From our analysis, we found a major limitation of the imaging model employed in conventional radiometric calibration methods and propose a new in-camera imaging model that fits well with today's cameras. With the new model, we present associated calibration procedures that allow us to convert sRGB images back to their original CCD RAW responses in a manner that is significantly more accurate than any existing methods. Additionally, we show how this new imaging model can be used to build an image correction application that converts an sRGB input image captured with the wrong camera settings to an sRGB output image that would have been recorded under the correct settings of a specific camera.",
"Photometric methods in computer vision require calibration of the camera's radiometric response, and previous works have addressed this problem using multiple registered images captured under different camera exposure settings. In many instances, such an image set is not available, so we propose a method that performs radiometric calibration from only a single image, based on measured RGB distributions at color edges. This technique automatically selects appropriate edge information for processing, and employs a Bayesian approach to compute the calibration. Extensive experimentation has shown that accurate calibration results can be obtained using only a single input image.",
"",
"We present a method of recovering high dynamic range radiance maps from photographs taken with conventional imaging equipment. In our method, multiple photographs of the scene are taken with different amounts of exposure. Our algorithm uses these differently exposed photographs to recover the response function of the imaging process, up to factor of scale, using the assumption of reciprocity. With the known response function, the algorithm can fuse the multiple photographs into a single, high dynamic range radiance map whose pixel values are proportional to the true radiance values in the scene. We demonstrate our method on images acquired with both photochemical and digital imaging processes. We discuss how this work is applicable in many areas of computer graphics involving digitized photographs, including image-based modeling, image compositing, and image processing. Lastly, we demonstrate a few applications of having high dynamic range radiance maps, such as synthesizing realistic motion blur and simulating the response of the human visual system.",
"Many vision applications require precise measurement of scene radiance. The function relating scene radiance to image intensity of an imaging system is called the camera response. We analyze the properties that all camera responses share. This allows us to find the constraints that any response function must satisfy. These constraints determine the theoretical space of all possible camera responses. We have collected a diverse database of real-world camera response functions (DoRF). Using this database, we show that real-world responses occupy a small part of the theoretical space of all possible responses. We combine the constraints from our theoretical space with the data from DoRF to create a low-parameter empirical model of response (EMoR). This response model allows us to accurately interpolate the complete response function of a camera from a small number of measurements obtained using a standard chart. We also show that the model can be used to accurately estimate the camera response from images of an arbitrary scene taken using different exposures. The DoRF database and the EMoR model can be downloaded at http: www.cs.columbia.edu CAVE.",
"In many computer vision systems, it is assumed that the image brightness of a point directly reflects the scene radiance of the point. However, the assumption does not hold in most cases due to nonlinear camera response function, exposure changes, and vignetting. The effects of these factors are most visible in image mosaics and textures of 3D models where colors look inconsistent and notable boundaries exist. In this paper, we propose a full radiometric calibration algorithm that includes robust estimation of the radiometric response function, exposures, and vignetting. By decoupling the effect of vignetting from the response function estimation, we approach each process in a manner that is robust to noise and outliers. We verify our algorithm with both synthetic and real data, which shows significant improvement compared to existing methods. We apply our estimation results to radiometrically align images for seamless mosaics and 3D model textures. We also use our method to create high dynamic range (HDR) mosaics that are more representative of the scene than normal mosaics."
]
} |
1707.08350 | 2740930075 | We present a novel deep learning framework that models the scene dependent image processing inside cameras. Often called as the radiometric calibration, the process of recovering RAW images from processed images (JPEG format in the sRGB color space) is essential for many computer vision tasks that rely on physically accurate radiance values. All previous works rely on the deterministic imaging model where the color transformation stays the same regardless of the scene and thus they can only be applied for images taken under the manual mode. In this paper, we propose a data-driven approach to learn the scene dependent and locally varying image processing inside cameras under the automode. Our method incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers. The results show that we can model the imaging pipeline of different cameras that operate under the automode accurately in both directions (from RAW to sRGB, from sRGB to RAW) and we show how we can apply our method to improve the performance of image deblurring. | In @cite_8 , a more complete in-camera imaging pipeline that includes processes such as the white balance, the color space transformation, and the gamut mapping in addition to the tone-mapping was introduced. With the new parametric model for the imaging, the work also presented an algorithm for recovering the parameters from a set of images using a scattered point interpolation scheme. Using a similar pipeline, a probabilistic model that takes into account the uncertainty in the color rendering was recently proposed in @cite_24 . | {
"cite_N": [
"@cite_24",
"@cite_8"
],
"mid": [
"2093580403",
"2127701282"
],
"abstract": [
"To produce images that are suitable for display, tone-mapping is widely used in digital cameras to map linear color measurements into narrow gamuts with limited dynamic range. This introduces non-linear distortion that must be undone, through a radiometric calibration process, before computer vision systems can analyze such photographs radiometrically. This paper considers the inherent uncertainty of undoing the effects of tone-mapping. We observe that this uncertainty varies substantially across color space, making some pixels more reliable than others. We introduce a model for this uncertainty and a method for fitting it to a given camera or imaging pipeline. Once fit, the model provides for each pixel in a tone-mapped digital photograph a probability distribution over linear scene colors that could have induced it. We demonstrate how these distributions can be useful for visual inference by incorporating them into estimation algorithms for a representative set of vision tasks.",
"We present a study of in-camera image processing through an extensive analysis of more than 10,000 images from over 30 cameras. The goal of this work is to investigate if image values can be transformed to physically meaningful values, and if so, when and how this can be done. From our analysis, we found a major limitation of the imaging model employed in conventional radiometric calibration methods and propose a new in-camera imaging model that fits well with today's cameras. With the new model, we present associated calibration procedures that allow us to convert sRGB images back to their original CCD RAW responses in a manner that is significantly more accurate than any existing methods. Additionally, we show how this new imaging model can be used to build an image correction application that converts an sRGB input image captured with the wrong camera settings to an sRGB output image that would have been recorded under the correct settings of a specific camera."
]
} |
1707.08350 | 2740930075 | We present a novel deep learning framework that models the scene dependent image processing inside cameras. Often called as the radiometric calibration, the process of recovering RAW images from processed images (JPEG format in the sRGB color space) is essential for many computer vision tasks that rely on physically accurate radiance values. All previous works rely on the deterministic imaging model where the color transformation stays the same regardless of the scene and thus they can only be applied for images taken under the manual mode. In this paper, we propose a data-driven approach to learn the scene dependent and locally varying image processing inside cameras under the automode. Our method incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers. The results show that we can model the imaging pipeline of different cameras that operate under the automode accurately in both directions (from RAW to sRGB, from sRGB to RAW) and we show how we can apply our method to improve the performance of image deblurring. | Recently, deep learning based image colorization has been studied @cite_11 @cite_1 , of which the objective is to restore chrominance information from a single channel luminance image. These works exploit the high-level semantic information to determine the chrominance of pixels by using CNNs, similar to those used in the high-level recognition tasks @cite_23 . In this paper, we show that color histogram based low-level features extracted using our deep network are more efficient for the given task compared to the high-level CNN features extracted from above previous work. | {
"cite_N": [
"@cite_1",
"@cite_23",
"@cite_11"
],
"mid": [
"2326925005",
"1686810756",
"2950064337"
],
"abstract": [
"Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result. The system is implemented as a feed-forward pass in a CNN at test time and is trained on over a million color images. We evaluate our algorithm using a “colorization Turing test,” asking human participants to choose between a generated and ground truth color image. Our method successfully fools humans on 32 of the trials, significantly higher than previous methods. Moreover, we show that colorization can be a powerful pretext task for self-supervised feature learning, acting as a cross-channel encoder. This approach results in state-of-the-art performance on several feature learning benchmarks.",
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
"We develop a fully automatic image colorization system. Our approach leverages recent advances in deep networks, exploiting both low-level and semantic representations. As many scene elements naturally appear according to multimodal color distributions, we train our model to predict per-pixel color histograms. This intermediate output can be used to automatically generate a color image, or further manipulated prior to image formation. On both fully and partially automatic colorization tasks, we outperform existing methods. We also explore colorization as a vehicle for self-supervised visual representation learning."
]
} |
1707.08350 | 2740930075 | We present a novel deep learning framework that models the scene dependent image processing inside cameras. Often called as the radiometric calibration, the process of recovering RAW images from processed images (JPEG format in the sRGB color space) is essential for many computer vision tasks that rely on physically accurate radiance values. All previous works rely on the deterministic imaging model where the color transformation stays the same regardless of the scene and thus they can only be applied for images taken under the manual mode. In this paper, we propose a data-driven approach to learn the scene dependent and locally varying image processing inside cameras under the automode. Our method incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers. The results show that we can model the imaging pipeline of different cameras that operate under the automode accurately in both directions (from RAW to sRGB, from sRGB to RAW) and we show how we can apply our method to improve the performance of image deblurring. | @cite_35 , an automatic photo adjustment method using a multi-layer perceptron was proposed. They feed the concatenation of global color features and semantic maps to a neural network system to find the scene dependent and the locally varying color mapping. As with the other data-driven image enhancement techniques @cite_31 @cite_15 , the features for the color mapping in their work are manually selected. However, one of the key properties behind the success of deep learning is in its ability to extract good features for the given task automatically. Instead of using manual features, we propose an end-to-end trainable deep neural network to model the scene dependent in-camera image processing. | {
"cite_N": [
"@cite_35",
"@cite_31",
"@cite_15"
],
"mid": [
"1920280450",
"2025328853",
"2149049515"
],
"abstract": [
"Photo retouching enables photographers to invoke dramatic visual impressions by artistically enhancing their photos through stylistic color and tone adjustments. However, it is also a time-consuming and challenging task that requires advanced skills beyond the abilities of casual photographers. Using an automated algorithm is an appealing alternative to manual work, but such an algorithm faces many hurdles. Many photographic styles rely on subtle adjustments that depend on the image content and even its semantics. Further, these adjustments are often spatially varying. Existing automatic algorithms are still limited and cover only a subset of these challenges. Recently, deep learning has shown unique abilities to address hard problems. This motivated us to explore the use of deep neural networks (DNNs) in the context of photo editing. In this article, we formulate automatic photo adjustment in a manner suitable for this approach. We also introduce an image descriptor accounting for the local semantics of an image. Our experiments demonstrate that training DNNs using these descriptors successfully capture sophisticated photographic styles. In particular and unlike previous techniques, it can model local adjustments that depend on image semantics. We show that this yields results that are qualitatively and quantitatively better than previous work.",
"Adjusting photographs to obtain compelling renditions requires skill and time. Even contrast and brightness adjustments are challenging because they require taking into account the image content. Photographers are also known for having different retouching preferences. As the result of this complexity, rule-based, one-size-fits-all automatic techniques often fail. This problem can greatly benefit from supervised machine learning but the lack of training data has impeded work in this area. Our first contribution is the creation of a high-quality reference dataset. We collected 5,000 photos, manually annotated them, and hired 5 trained photographers to retouch each picture. The result is a collection of 5 sets of 5,000 example input-output pairs that enable supervised learning. We first use this dataset to predict a user's adjustment from a large training set. We then show that our dataset and features enable the accurate adjustment personalization using a carefully chosen set of training photos. Finally, we introduce difference learning: this method models and predicts difference between users. It frees the user from using predetermined photos for training. We show that difference learning enables accurate prediction using only a handful of examples.",
"In this paper, we describe a technique to automatically enhance the perceptual quality of an image. Unlike previous techniques, where global statistics of the image are used to determine enhancement operation, our method is local and relies on local scene descriptors and context in addition to high-level image statistics. We cast the problem of image enhancement as searching for the best transformation for each pixel in the given image and then discovering the enhanced image using a formulation based on Gaussian Random Fields. The search is done in a coarse-to-fine manner, namely by finding the best candidate images, followed by pixels. Our experiments indicate that such context-based local enhancement is better than global enhancement schemes. A user study using Mechanical Turk shows that the subjects prefer contextual and local enhancements over the ones provided by existing schemes."
]
} |
1707.08557 | 2736373710 | This work presents a new method to quantify connectivity in transportation networks. Inspired by the field of topological data analysis, we propose a novel approach to explore the robustness of road network connectivity in the presence of congestion on the roadway. The robustness of the pattern is summarized in a congestion barcode, which can be constructed directly from traffic datasets commonly used for navigation. As an initial demonstration, we illustrate the main technique on a publicly available traffic dataset in a neighborhood in New York City. | A variety of topological information can be studied. The simplest homology" counts the number of connected components in the topology. One can also construct homologies reflecting the structure of ; this has implications for robotic path planning in the face of uncertainty @cite_8 and also connectivity in the brain @cite_22 . By way of precedence, the article @cite_1 , uses persistent homology to measure similarities between road maps. | {
"cite_N": [
"@cite_1",
"@cite_22",
"@cite_8"
],
"mid": [
"2030523990",
"1967760180",
"2037034267"
],
"abstract": [
"We define a topology-based distance metric between road networks embedded in the plane. This distance measure is based on local persistent homology, and employs a local distance signature that enables identification and visualization of local differences between the road networks. This paper is motivated by the need to recognize changes in road networks over time and to assess the quality of different map construction algorithms. One particular challenge is evaluating the results when no ground truth is known. However, we demonstrate that we can overcome this hurdle by using a statistical technique known as the bootstrap.",
"The brain network is usually constructed by estimating the connectivity matrix and thresholding it at an arbitrary level. The problem with this standard method is that we do not have any generally accepted criteria for determining a proper threshold. Thus, we propose a novel multiscale framework that models all brain networks generated over every possible threshold. Our approach is based on persistent homology and its various representations such as the Rips filtration, barcodes, and dendrograms. This new persistent homological framework enables us to quantify various persistent topological features at different scales in a coherent manner. The barcode is used to quantify and visualize the evolutionary changes of topological features such as the Betti numbers over different scales. By incorporating additional geometric information to the barcode, we obtain a single linkage dendrogram that shows the overall evolution of the network. The difference between the two networks is then measured by the Gromov-Hausdorff distance over the dendrograms. As an illustration, we modeled and differentiated the FDG-PET based functional brain networks of 24 attention-deficit hyperactivity disorder children, 26 autism spectrum disorder children, and 11 pediatric control subjects.",
"We address the fundamental problem of goal-directed path planning in an uncertain environment represented as a probability (of occupancy) map. Most methods generally use a threshold to reduce the grayscale map to a binary map before applying off-the-shelf techniques to find the best path. This raises the somewhat ill-posed question, what is the right (optimal) value to threshold the map? We instead suggest a persistent homology approach to the problem-a topological approach in which we seek the homology class of trajectories that is most persistent for the given probability map. In other words, we want the class of trajectories that is free of obstacles over the largest range of threshold values. In order to make this problem tractable, we use homology in ℤ 2 coefficients (instead of the standard ℤ coefficients), and describe how graph search-based algorithms can be used to find trajectories in different homology classes. Our simulation results demonstrate the efficiency and practical applicability of the algorithm proposed in this paper.paper."
]
} |
1707.08347 | 2736348570 | We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image quality is known. These ranked image sets can be automatically generated without laborious human labeling. We then use fine-tuning to transfer the knowledge represented in the trained Siamese Network to a traditional CNN that estimates absolute image quality from single images. We demonstrate how our approach can be made significantly more efficient than traditional Siamese Networks by forward propagating a batch of images through a single network and backpropagating gradients derived from all pairs of images in the batch. Experiments on the TID2013 benchmark show that we improve the state-of-the-art by over 5 . Furthermore, on the LIVE benchmark we show that our approach is superior to existing NR-IQA techniques and that we even outperform the state-of-the-art in full-reference IQA (FR-IQA) methods without having to resort to high-quality reference images to infer IQA. | Traditional NR-IQA approaches. Most traditional NR-IQA can be classified into Natural Scene Statistics (NSS) methods and learning-based methods. In NSS methods, the assumption is that images of different quality vary in the statistics of responses to specific filters. Wavelets @cite_24 , DCT @cite_33 and Curvelets @cite_3 are commonly used to extract the features in different sub-bands. These feature distributions are parametrized, for example with the Generalized Gaussian Distribution @cite_12 . The aim of these methods is to estimate the distributional parameters, from which a quality assessment can be inferred. The authors of @cite_4 propose to extract NSS features in the spatial domain to obtain significant speed-ups. In learning-based methods, local features are extracted and mapped to the MOS using, for example, Support Machine Regression or Neural Networks @cite_41 . The codebook method @cite_37 @cite_17 combines different features instead of using local features directly. Datasets without MOS can be exploited to construct the codebook by means of unsupervised learning, which is particularly important due to of the small size of existing datasets. Saliency maps @cite_32 can be used to model human vision system and improve precision in these methods. | {
"cite_N": [
"@cite_37",
"@cite_4",
"@cite_33",
"@cite_41",
"@cite_32",
"@cite_3",
"@cite_24",
"@cite_12",
"@cite_17"
],
"mid": [
"1987489060",
"1982471090",
"2162692770",
"2804524105",
"1922292441",
"2008051965",
"2163370434",
"",
"2061513831"
],
"abstract": [
"In this paper, we present an efficient general-purpose objective no-reference (NR) image quality assessment (IQA) framework based on unsupervised feature learning. The goal is to build a computational model to automatically predict human perceived image quality without a reference image and without knowing the distortion present in the image. Previous approaches for this problem typically rely on hand-crafted features which are carefully designed based on prior knowledge. In contrast, we use raw-image-patches extracted from a set of unlabeled images to learn a dictionary in an unsupervised manner. We use soft-assignment coding with max pooling to obtain effective image representations for quality estimation. The proposed algorithm is very computationally appealing, using raw image patches as local descriptors and using soft-assignment for encoding. Furthermore, unlike previous methods, our unsupervised feature learning strategy enables our method to adapt to different domains. CORNIA (Codebook Representation for No-Reference Image Assessment) is tested on LIVE database and shown to perform statistically better than the full-reference quality measure, structural similarity index (SSIM) and is shown to be comparable to state-of-the-art general purpose NR-IQA algorithms.",
"We propose a natural scene statistic-based distortion-generic blind no-reference (NR) image quality assessment (IQA) model that operates in the spatial domain. The new model, dubbed blind referenceless image spatial quality evaluator (BRISQUE) does not compute distortion-specific features, such as ringing, blur, or blocking, but instead uses scene statistics of locally normalized luminance coefficients to quantify possible losses of “naturalness” in the image due to the presence of distortions, thereby leading to a holistic measure of quality. The underlying features used derive from the empirical distribution of locally normalized luminances and products of locally normalized luminances under a spatial natural scene statistic model. No transformation to another coordinate frame (DCT, wavelet, etc.) is required, distinguishing it from prior NR IQA approaches. Despite its simplicity, we are able to show that BRISQUE is statistically better than the full-reference peak signal-to-noise ratio and the structural similarity index, and is highly competitive with respect to all present-day distortion-generic NR IQA algorithms. BRISQUE has very low computational complexity, making it well suited for real time applications. BRISQUE features may be used for distortion-identification as well. To illustrate a new practical application of BRISQUE, we describe how a nonblind image denoising algorithm can be augmented with BRISQUE in order to perform blind image denoising. Results show that BRISQUE augmentation leads to performance improvements over state-of-the-art methods. A software release of BRISQUE is available online: http: live.ece.utexas.edu research quality BRISQUE_release.zip for public use and evaluation.",
"We develop an efficient general-purpose blind no-reference image quality assessment (IQA) algorithm using a natural scene statistics (NSS) model of discrete cosine transform (DCT) coefficients. The algorithm is computationally appealing, given the availability of platforms optimized for DCT computation. The approach relies on a simple Bayesian inference model to predict image quality scores given certain extracted features. The features are based on an NSS model of the image DCT coefficients. The estimated parameters of the model are utilized to form features that are indicative of perceptual quality. These features are used in a simple Bayesian inference approach to predict quality scores. The resulting algorithm, which we name BLIINDS-II, requires minimal training and adopts a simple probabilistic model for score prediction. Given the extracted features from a test image, the quality score that maximizes the probability of the empirically determined inference model is chosen as the predicted quality score of that image. When tested on the LIVE IQA database, BLIINDS-II is shown to correlate highly with human judgments of quality, at a level that is competitive with the popular SSIM index.",
"",
"Image quality assessment (IQA) tries to estimate human perception based image visual quality in an objective manner. Existing approaches target this problem with or without reference images. For no-reference image quality assessment, there is no given reference image or any knowledge of the distortion type of the image. Previous approaches measure the image quality from signal level rather than semantic analysis. They typically depend on various features to represent local characteristic of an image. In this paper we propose a new no-reference (NR) image quality assessment (IQA) framework based on semantic obviousness. We discover that semantic-level factors affect human perception of image quality. With such observation, we explore semantic obviousness as a metric to perceive objects of an image. We propose to extract two types of features, one to measure the semantic obviousness of the image and the other to discover local characteristic. Then the two kinds of features are combined for image quality estimation. The principles proposed in our approach can also be incorporated with many existing IQA algorithms to boost their performance. We evaluate our approach on the LIVE dataset. Our approach is demonstrated to be superior to the existing NR-IQA algorithms and comparable to the state-of-the-art full-reference IQA (FR-IQA) methods. Cross-dataset experiments show the generalization ability of our approach.",
"Abstract We study the efficacy of utilizing a powerful image descriptor, the curvelet transform, to learn a no-reference (NR) image quality assessment (IQA) model. A set of statistical features are extracted from a computed image curvelet representation, including the coordinates of the maxima of the log-histograms of the curvelet coefficients values, and the energy distributions of both orientation and scale in the curvelet domain. Our results indicate that these features are sensitive to the presence and severity of image distortion. Operating within a 2-stage framework of distortion classification followed by quality assessment, we train an image distortion and quality prediction engine using a support vector machine (SVM). The resulting algorithm, dubbed CurveletQA for short, was tested on the LIVE IQA database and compared to state-of-the-art NR FR IQA algorithms. We found that CurveletQA correlates well with human subjective opinions of image quality, delivering performance that is competitive with popular full-reference (FR) IQA algorithms such as SSIM, and with top-performing NR IQA models. At the same time, CurveletQA has a relatively low complexity.",
"Present day no-reference no-reference image quality assessment (NR IQA) algorithms usually assume that the distortion affecting the image is known. This is a limiting assumption for practical applications, since in a majority of cases the distortions in the image are unknown. We propose a new two-step framework for no-reference image quality assessment based on natural scene statistics (NSS). Once trained, the framework does not require any knowledge of the distorting process and the framework is modular in that it can be extended to any number of distortions. We describe the framework for blind image quality assessment and a version of this framework-the blind image quality index (BIQI) is evaluated on the LIVE image quality assessment database. A software release of BIQI has been made available online: http: live.ece.utexas.edu research quality BIQI_release.zip.",
"",
"The goal of no-reference objective image quality assessment (NR-IQA) is to develop a computational model that can predict the human-perceived quality of distorted images accurately and automatically without any prior knowledge of reference images. Most existing NR-IQA approaches are distortion specific and are typically limited to one or two specific types of distortions. In most practical applications, however, information about the distortion type is not really available. In this paper, we propose a general-purpose NR-IQA approach based on visual codebooks. A visual codebook consisting of Gabor-filter-based local features extracted from local image patches is used to capture complex statistics of a natural image. The codebook encodes statistics by quantizing the feature space and accumulating histograms of patch appearances. This method does not assume any specific types of distortions; however, when evaluating images with a particular type of distortion, it does require examples with the same or similar distortion for training. Experimental results demonstrate that the predicted quality score using our method is consistent with human-perceived image quality. The proposed method is comparable to state-of-the-art general-purpose NR-IQA methods and outperforms the full-reference image quality metrics, peak signal-to-noise ratio and structural similarity index on the Laboratory for Image and Video Engineering IQA database."
]
} |
1707.08347 | 2736348570 | We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image quality is known. These ranked image sets can be automatically generated without laborious human labeling. We then use fine-tuning to transfer the knowledge represented in the trained Siamese Network to a traditional CNN that estimates absolute image quality from single images. We demonstrate how our approach can be made significantly more efficient than traditional Siamese Networks by forward propagating a batch of images through a single network and backpropagating gradients derived from all pairs of images in the batch. Experiments on the TID2013 benchmark show that we improve the state-of-the-art by over 5 . Furthermore, on the LIVE benchmark we show that our approach is superior to existing NR-IQA techniques and that we even outperform the state-of-the-art in full-reference IQA (FR-IQA) methods without having to resort to high-quality reference images to infer IQA. | Deep learning for NR-IQA. In recent years several works have used deep learning for NR-IQA @cite_8 @cite_31 @cite_30 . One of the main drawbacks of deep networks is the need for large labeled datasets, which are currently not available for NR-IQA research. To address this problem @cite_31 consider small @math patches rather than images, thereby greatly augmenting the number of training examples. The authors of @cite_20 @cite_30 follow the same pipeline. @cite_30 the authors design a multi-task CNN to learn the type of distortions and image quality simultaneously. Bianco at al. @cite_8 propose to use a pre-trained network to mitigate the lack of training data. They extract features from a pre-trained model fine-tuned on an IQA dataset. These features are then used to train an SVR model to map features to IQA scores. | {
"cite_N": [
"@cite_30",
"@cite_31",
"@cite_20",
"@cite_8"
],
"mid": [
"2294857031",
"2051596736",
"2509123681",
"2286686646"
],
"abstract": [
"",
"In this work we describe a Convolutional Neural Network (CNN) to accurately predict image quality without a reference image. Taking image patches as input, the CNN works in the spatial domain without using hand-crafted features that are employed by most previous methods. The network consists of one convolutional layer with max and min pooling, two fully connected layers and an output node. Within the network structure, feature learning and regression are integrated into one optimization process, which leads to a more effective model for estimating image quality. This approach achieves state of the art performance on the LIVE dataset and shows excellent generalization ability in cross dataset experiments. Further experiments on images with local distortions demonstrate the local quality estimation ability of our CNN, which is rarely reported in previous literature.",
"This paper presents a no reference image (NR) quality assessment (IQA) method based on a deep convolutional neural network (CNN). The CNN takes unpreprocessed image patches as an input and estimates the quality without employing any domain knowledge. By that, features and natural scene statistics are learnt purely data driven and combined with pooling and regression in one framework. We evaluate the network on the LIVE database and achieve a linear Pearson correlation superior to state-of-the-art NR IQA methods. We also apply the network to the image forensics task of decoder-sided quantization parameter estimation and also here achieve correlations of r = 0.989.",
"In this work, we investigate the use of deep learning for distortion-generic blind image quality assessment. We report on different design choices, ranging from the use of features extracted from pre-trained convolutional neural networks (CNNs) as a generic image description, to the use of features extracted from a CNN fine-tuned for the image quality task. Our best proposal, named DeepBIQ, estimates the image quality by average-pooling the scores predicted on multiple subregions of the original image. Experimental results on the LIVE In the Wild Image Quality Challenge Database show that DeepBIQ outperforms the state-of-the-art methods compared, having a linear correlation coefficient with human subjective scores of almost 0.91. These results are further confirmed also on four benchmark databases of synthetically distorted images: LIVE, CSIQ, TID2008, and TID2013."
]
} |
1707.08347 | 2736348570 | We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image quality is known. These ranked image sets can be automatically generated without laborious human labeling. We then use fine-tuning to transfer the knowledge represented in the trained Siamese Network to a traditional CNN that estimates absolute image quality from single images. We demonstrate how our approach can be made significantly more efficient than traditional Siamese Networks by forward propagating a batch of images through a single network and backpropagating gradients derived from all pairs of images in the batch. Experiments on the TID2013 benchmark show that we improve the state-of-the-art by over 5 . Furthermore, on the LIVE benchmark we show that our approach is superior to existing NR-IQA techniques and that we even outperform the state-of-the-art in full-reference IQA (FR-IQA) methods without having to resort to high-quality reference images to infer IQA. | Hard-negative mining for Siamese network training. It is well known that a naive approach to sampling pairs to training Siamese networks is suboptimal. To address this problem several approaches to hard-negative mining have been proposed. @cite_42 , they propose a hard positive and hard-negative mining strategy to forward-propagate a set of pairs and sample the highest loss pairs with back-propagation. However, hard mining comes with a high computational cost (they report an increase of up to 80 ). @cite_25 they propose semi-hard pair selection, arguing that selecting hardest pairs can lead to bad local minima. The batch size used is around 1800 examples, which again leads to a considerable increase in computational cost. @cite_23 the authors take a batch of pairs as input and choose the four hardest negative samples within the mini-batch. To solve for a bad local optimum, @cite_43 optimize a smooth upper bound loss function to take advantage of all possible pairs in the mini-batch. | {
"cite_N": [
"@cite_43",
"@cite_42",
"@cite_25",
"@cite_23"
],
"mid": [
"2176040302",
"1869500417",
"2096733369",
"219040644"
],
"abstract": [
"Learning the distance metric between pairs of examples is of great importance for learning and visual recognition. With the remarkable success from the state of the art convolutional neural networks, recent works have shown promising results on discriminatively training the networks to learn semantic feature embeddings where similar examples are mapped close to each other and dissimilar examples are mapped farther apart. In this paper, we describe an algorithm for taking full advantage of the training batches in the neural network training by lifting the vector of pairwise distances within the batch to the matrix of pairwise distances. This step enables the algorithm to learn the state of the art feature embedding by optimizing a novel structured prediction objective on the lifted problem. Additionally, we collected Online Products dataset: 120k images of 23k classes of online products for metric learning. Our experiments on the CUB-200-2011, CARS196, and Online Products datasets demonstrate significant improvement over existing deep feature embedding methods on all experimented embedding sizes with the GoogLeNet network.",
"Deep learning has revolutionalized image-level tasks such as classification, but patch-level tasks, such as correspondence, still rely on hand-crafted features, e.g. SIFT. In this paper we use Convolutional Neural Networks (CNNs) to learn discriminant patch representations and in particular train a Siamese network with pairs of (non-)corresponding patches. We deal with the large number of potential pairs with the combination of a stochastic sampling of the training set and an aggressive mining strategy biased towards patches that are hard to classify. By using the L2 distance during both training and testing we develop 128-D descriptors whose euclidean distances reflect patch similarity, and which can be used as a drop-in replacement for any task involving SIFT. We demonstrate consistent performance gains over the state of the art, and generalize well against scaling and rotation, perspective transformation, non-rigid deformation, and illumination changes. Our descriptors are efficient to compute and amenable to modern GPUs, and are publicly available.",
"Despite significant recent advances in the field of face recognition [10, 14, 15, 17], implementing face verification and recognition efficiently at scale presents serious challenges to current approaches. In this paper we present a system, called FaceNet, that directly learns a mapping from face images to a compact Euclidean space where distances directly correspond to a measure of face similarity. Once this space has been produced, tasks such as face recognition, verification and clustering can be easily implemented using standard techniques with FaceNet embeddings as feature vectors.",
"Is strong supervision necessary for learning a good visual representation? Do we really need millions of semantically-labeled images to train a Convolutional Neural Network (CNN)? In this paper, we present a simple yet surprisingly powerful approach for unsupervised learning of CNN. Specifically, we use hundreds of thousands of unlabeled videos from the web to learn visual representations. Our key idea is that visual tracking provides the supervision. That is, two patches connected by a track should have similar visual representation in deep feature space since they probably belong to same object or object part. We design a Siamese-triplet network with a ranking loss function to train this CNN representation. Without using a single image from ImageNet, just using 100K unlabeled videos and the VOC 2012 dataset, we train an ensemble of unsupervised networks that achieves 52 mAP (no bounding box regression). This performance comes tantalizingly close to its ImageNet-supervised counterpart, an ensemble which achieves a mAP of 54.4 . We also show that our unsupervised network can perform competitively in other tasks such as surface-normal estimation."
]
} |
1707.08186 | 2738789134 | In [SPAA2007], define a streaming B-tree (or index) as one that supports updates in amortized @math IOs, and present a structure achieving amortized @math IOs and queries in @math IOs. We extend their result to the partially-persistent case. For a version @math , let @math be the number of keys accessible at @math and @math be the total number of updates. We give a data structure using space @math , supporting updates to a leaf version @math with @math amortized IOs and answering range queries returning @math elements with @math IOs on average (where the average is over all queries covering disjoint key ranges at a given version). This is the first persistent streaming' index we are aware of, i.e. that supports updates in @math IOs and supports efficient range queries. | Becker @cite_10 presented the multiversion B-tree (MVBT), which solves the space blowup problem. It achieves the same query and update bounds with @math space, and is partially-persistent. 115861 developed two fully-persistent B-tree variants based on variants of the techniques from @cite_6 , but for both variants, either there is a large space blowup, or range queries may be far from optimal. Recently, @cite_3 presented a fully-persistent B-tree that uses linear space, and achieves roughly the same update query bounds (within a small factor) as the MVBT and the CoW B-tree. These results are summarized in Table . | {
"cite_N": [
"@cite_3",
"@cite_10",
"@cite_6"
],
"mid": [
"",
"2055385473",
"2142947709"
],
"abstract": [
"",
"In a variety of applications, we need to keep track of the development of a data set over time. For maintaining and querying these multiversion data efficiently, external storage structures are an absolute necessity. We propose a multiversion B-tree that supports insertions and deletions of data items at the current version and range queries and exact match queries for any version, current or past. Our multiversion B-tree is asymptotically optimal in the sense that the time and space bounds are asymptotically the same as those of the (single-version) B-tree in the worst case. The technique we present for transforming a (single-version) B-tree into a multiversion B-tree is quite general: it applies to a number of hierarchical external access structures with certain properties directly, and it can be modified for others.",
"Abstract This paper is a study of persistence in data structures. Ordinary data structures are ephemeral in the sense that a change to the structure destroys the old version, leaving only the new version available for use. In contrast, a persistent structure allows access to any version, old or new, at any time. We develop simple, systematic, and efficient techniques for making linked data structures persistent. We use our techniques to devise persistent forms of binary search trees with logarithmic access, insertion, and deletion times and O (1) space bounds for insertion and deletion."
]
} |
1707.08186 | 2738789134 | In [SPAA2007], define a streaming B-tree (or index) as one that supports updates in amortized @math IOs, and present a structure achieving amortized @math IOs and queries in @math IOs. We extend their result to the partially-persistent case. For a version @math , let @math be the number of keys accessible at @math and @math be the total number of updates. We give a data structure using space @math , supporting updates to a leaf version @math with @math amortized IOs and answering range queries returning @math elements with @math IOs on average (where the average is over all queries covering disjoint key ranges at a given version). This is the first persistent streaming' index we are aware of, i.e. that supports updates in @math IOs and supports efficient range queries. | @cite_0 prove that any partially-persistent CO index that answers queries in @math IOs must use superlinear space. Our query bound comes close to this, except that the first term in our bound is @math , and the bound is an average-case bound, not worst-case. It remains open to improve our query bound to a worst-case one. | {
"cite_N": [
"@cite_0"
],
"mid": [
"2059543065"
],
"abstract": [
"We consider a number of range reporting problems in two and three dimensions and prove lower bounds on the amount of space required by any cache-oblivious data structure for these problems that achieves an optimal query bound of O(logBN + K B) block transfers in the worst case, where K is the size of the query output. The problems we study are three-sided range reporting, 3-d dominance reporting, and 3-d halfspace range reporting. We prove that, in order to achieve the above query bound or even a bound of O((logB N)c (1 + K B)), for any constant c > 0, the structure has to use Ohmega(N (log log N)e) space, where e > 0 is a constant that depends on c and on the constant hidden in the big-Oh notation of the query bound. Our result has a number of interesting consequences. The first one is a new type of separation between the I O model and the cache-oblivious model, as I O-efficient data structures with the optimal query bound and using linear or O(N logAST N) space are known for the above problems. The second consequence is the non-existence of a linear-space cache-oblivious persistent B-tree with worst-case optimal 1-d range reporting queries."
]
} |
1707.08525 | 2736616773 | Identification and counting of cells and mitotic figures is a standard task in diagnostic histopathology. Due to the large overall cell count on histological slides and the potential sparse prevalence of some relevant cell types or mitotic figures, retrieving annotation data for sufficient statistics is a tedious task and prone to a significant error in assessment. Automatic classification and segmentation is a classic task in digital pathology, yet it is not solved to a sufficient degree. We present a novel approach for cell and mitotic figure classification, based on a deep convolutional network with an incorporated Spatial Transformer Network. The network was trained on a novel data set with ten thousand mitotic figures, about ten times more than previous data sets. The algorithm is able to derive the cell class (mitotic tumor cells, non-mitotic tumor cells and granulocytes) and their position within an image. The mean accuracy of the algorithm in a five-fold cross-validation is 91.45 . In our view, the approach is a promising step into the direction of a more objective and accurate, semi-automatized mitosis counting supporting the pathologist. | Typically, the process of object detection is parted into two sub-processes: Segmentation and classification. This setup is especially sensible for histology since the images represent a large amount of data and classification is usually the more complex process compared to segmentation. Sommer used pixel-wise classification for candidate retrieval and then object shape and texture features for mitotic cell classification @cite_2 . Irshad used active contour models for candidate selection and statistical and morphological features for classification @cite_11 . Those hand-crafted features have significant drawbacks, however: Given the often small data sets, automatic selection of features is prone to random correlation, while using higher-dimensional classification approaches on the complete set increases overfitting @cite_1 . Further, it is questionable, if those approaches can represent the variability in shape and texture of mitotic figures @cite_9 . | {
"cite_N": [
"@cite_9",
"@cite_11",
"@cite_1",
"@cite_2"
],
"mid": [
"2560920277",
"2026893473",
"",
"1792334496"
],
"abstract": [
"The number of mitoses per tissue area gives an important aggressiveness indication of the invasive breast carcinoma. However, automatic mitosis detection in histology images remains a challenging problem. Traditional methods either employ hand-crafted features to discriminate mitoses from other cells or construct a pixel-wise classifier to label every pixel in a sliding window way. While the former suffers from the large shape variation of mitoses and the existence of many mimics with similar appearance, the slow speed of the later prohibits its use in clinical practice. In order to overcome these shortcomings, we propose a fast and accurate method to detect mitosis by designing a novel deep cascaded convolutional neural network, which is composed of two components. First, by leveraging the fully convolutional neural network, we propose a coarse retrieval model to identify and locate the candidates of mitosis while preserving a high sensitivity. Based on these candidates, a fine discrimination model utilizing knowledge transferred from cross-domain is developed to further single out mitoses from hard mimics. Our approach outperformed other methods by a large margin in 2014 ICPR MITOS-ATYPIA challenge in terms of detection accuracy. When compared with the state-of-the-art methods on the 2012 ICPR MITOSIS data (a smaller and less challenging dataset), our method achieved comparable or better results with a roughly 60 times faster speed.",
"Context: According to Nottingham grading system, mitosis count plays a critical role in cancer diagnosis and grading. Manual counting of mitosis is tedious and subject to considerable inter- and intra-reader variations. Aims: The aim is to improve the accuracy of mitosis detection by selecting the color channels that better capture the statistical and morphological features, which classify mitosis from other objects. Materials and Methods: We propose a framework that includes comprehensive analysis of statistics and morphological features in selected channels of various color spaces that assist pathologists in mitosis detection. In candidate detection phase, we perform Laplacian of Gaussian, thresholding, morphology and active contour model on blue-ratio image to detect and segment candidates. In candidate classification phase, we extract a total of 143 features including morphological, first order and second order (texture) statistics features for each candidate in selected channels and finally classify using decision tree classifier. Results and Discussion: The proposed method has been evaluated on Mitosis Detection in Breast Cancer Histological Images (MITOS) dataset provided for an International Conference on Pattern Recognition 2012 contest and achieved 74 and 71 detection rate, 70 and 56 precision and 72 and 63 F-Measure on Aperio and Hamamatsu images, respectively. Conclusions and Future Work: The proposed multi-channel features computation scheme uses fixed image scale and extracts nuclei features in selected channels of various color spaces. This simple but robust model has proven to be highly efficient in capturing multi-channels statistical features for mitosis detection, during the MITOS international benchmark. Indeed, the mitosis detection of critical importance in cancer diagnosis is a very challenging visual task. In future work, we plan to use color deconvolution as preprocessing and Hough transform or local extrema based candidate detection in order to reduce the number of candidates in mitosis and non-mitosis classes.",
"",
"Breast cancer grading of histological tissue samples by visual inspection is the standard clinical practice for the diagnosis and prognosis of cancer development. An important parameter for tumor prognosis is the number of mitotic cells present in histologically stained breast cancer tissue sections. We propose a hierarchical learning workflow for automated mitosis detection in breast cancer. From an initial training set a pixel-wise classifier is learned to segment candidate cells, which are then classified into mitotic and non-mitotic cells using object shape and texture features. Our workflow banks on two open source biomedical image analysis software: “ilastik” and “CellCognition” which provide a user user friendly interface to powerful learning algorithms, with the potential of making the pathologist work an easier task. We evaluate our approach on a dataset of 35 highresolution histopathological images from 5 different specimen (provided by International Conference for Pattern Recognition 2012 contest on Mitosis Detection in Breast Cancer Histological Images). Based on the candidate segmentation our approach achieves an area-under Precision-Recall-curve of 70 on an annotated dataset, with good localization accuracy, little parameter tuning and small user effort. Source code is provided."
]
} |
1707.08470 | 2460515658 | Over the years, Twitter has become one of the largest communication platforms providing key data to various applications such as brand monitoring, trend detection, among others. Entity linking is one of the major tasks in natural language understanding from tweets and it associates entity mentions in text to corresponding entries in knowledge bases in order to provide unambiguous interpretation and additional context. State-of-the-art techniques have focused on linking explicitly mentioned entities in tweets with reasonable success. However, we argue that in addition to explicit mentions --- i.e. 'The movie Gravity was more expensive than the mars orbiter mission' --- entities movie Gravity can also be mentioned implicitly --- i.e. 'This new space movie is crazy. you must watch it!.' This paper introduces the problem of implicit entity linking in tweets. We propose an approach that models the entities by exploiting their factual and contextual knowledge. We demonstrate how to use these models to perform implicit entity linking on a ground truth dataset with 397 tweets from two domains, namely, Movie and Book. Specifically, we show: 1 the importance of linking implicit entities and its value addition to the standard entity linking task, and 2 the importance of exploiting contextual knowledge associated with an entity for linking their implicit mentions. We also make the ground truth dataset publicly available to foster the research in this new research areacity. | Meij, et al @cite_15 derives the topics of a tweet. They extract features from the tweet and the Wikipedia pages of entities, and apply machine learning algorithms to derive the topic. We found that this work has focused on deriving topic using explicit entities as their evaluation dataset contains only 16 tweets whose label of the manually annotated topic is not present in the tweet text (i.e. not a string match). Nevertheless, they are found to be either synonyms or related entities to the explicit entities in the tweets and not implicit entity mentions (e.g. New York and Big Apple, Almighty and God, stuffy nose and Rhinitis). | {
"cite_N": [
"@cite_15"
],
"mid": [
"2013579020"
],
"abstract": [
"Microblogs have become an important source of information for the purpose of marketing, intelligence, and reputation management. Streams of microblogs are of great value because of their direct and real-time nature. Determining what an individual microblog post is about, however, can be non-trivial because of creative language usage, the highly contextualized and informal nature of microblog posts, and the limited length of this form of communication. We propose a solution to the problem of determining what a microblog post is about through semantic linking: we add semantics to posts by automatically identifying concepts that are semantically related to it and generating links to the corresponding Wikipedia articles. The identified concepts can subsequently be used for, e.g., social media mining, thereby reducing the need for manual inspection and selection. Using a purpose-built test collection of tweets, we show that recently proposed approaches for semantic linking do not perform well, mainly due to the idiosyncratic nature of microblog posts. We propose a novel method based on machine learning with a set of innovative features and show that it is able to achieve significant improvements over all other methods, especially in terms of precision."
]
} |
1707.08470 | 2460515658 | Over the years, Twitter has become one of the largest communication platforms providing key data to various applications such as brand monitoring, trend detection, among others. Entity linking is one of the major tasks in natural language understanding from tweets and it associates entity mentions in text to corresponding entries in knowledge bases in order to provide unambiguous interpretation and additional context. State-of-the-art techniques have focused on linking explicitly mentioned entities in tweets with reasonable success. However, we argue that in addition to explicit mentions --- i.e. 'The movie Gravity was more expensive than the mars orbiter mission' --- entities movie Gravity can also be mentioned implicitly --- i.e. 'This new space movie is crazy. you must watch it!.' This paper introduces the problem of implicit entity linking in tweets. We propose an approach that models the entities by exploiting their factual and contextual knowledge. We demonstrate how to use these models to perform implicit entity linking on a ground truth dataset with 397 tweets from two domains, namely, Movie and Book. Specifically, we show: 1 the importance of linking implicit entities and its value addition to the standard entity linking task, and 2 the importance of exploiting contextual knowledge associated with an entity for linking their implicit mentions. We also make the ground truth dataset publicly available to foster the research in this new research areacity. | The word-level tweet linking has two main steps: The word-level entity linking has been studied extensively for text like Wikipedia and News @cite_10 @cite_12 @cite_4 @cite_1 @cite_7 ; however, these approaches have proved to be ineffective on short and noisy text like tweets @cite_2 @cite_8 . Here we will discuss the approaches taken to solve this problem in tweets. The first step was performed by matching the word sequences of the tweet to the page titles and the anchor texts in Wikipedia and consider all matching pages and pages redirected by matching anchor texts to be candidates @cite_11 @cite_8 @cite_6 . The second step was performed by optimizing the relatedness calculated among the candidate entities @cite_8 @cite_5 or based on the threshold defined over measures that calculate the similarity between the entity mention and entity representation @cite_11 , or by applying structural learning techniques @cite_6 . An approach to solve the implicit entity linking in tweets has to take a fresh perspective since by definition it has neither anchor text nor the page title present in the tweet. | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_1",
"@cite_6",
"@cite_2",
"@cite_5",
"@cite_10",
"@cite_12",
"@cite_11"
],
"mid": [
"11298561",
"2100341149",
"2123142779",
"2104583100",
"2139694477",
"2143933463",
"2252250824",
"86887328",
"1541691357",
"2396582863"
],
"abstract": [
"Disambiguating named entities in natural-language text maps mentions of ambiguous names onto canonical entities like people or places, registered in a knowledge base such as DBpedia or YAGO. This paper presents a robust method for collective disambiguation, by harnessing context from knowledge bases and using a new form of coherence graph. It unifies prior approaches into a comprehensive framework that combines three measures: the prior probability of an entity being mentioned, the similarity between the contexts of a mention and a candidate entity, as well as the coherence among candidate entities for all mentions together. The method builds a weighted graph of mentions and candidate entities, and computes a dense subgraph that approximates the best joint mention-entity mapping. Experiments show that the new method significantly outperforms prior methods in terms of accuracy, with robust behavior across a variety of inputs.",
"This paper describes how to automatically cross-reference documents with Wikipedia: the largest knowledge base ever known. It explains how machine learning can be used to identify significant terms within unstructured text, and enrich it with links to the appropriate Wikipedia articles. The resulting link detector and disambiguator performs very well, with recall and precision of almost 75 . This performance is constant whether the system is evaluated on Wikipedia articles or \"real world\" documents. This work has implications far beyond enriching documents with explanatory links. It can provide structured knowledge about any unstructured fragment of text. Any task that is currently addressed with bags of words - indexing, clustering, retrieval, and summarization to name a few - could use the techniques described here to draw on a vast network of concepts and semantics.",
"We designed and implemented TAGME, a system that is able to efficiently and judiciously augment a plain-text with pertinent hyperlinks to Wikipedia pages. The specialty of TAGME with respect to known systems [5,8] is that it may annotate texts which are short and poorly composed, such as snippets of search-engine results, tweets, news, etc.. This annotation is extremely informative, so any task that is currently addressed using the bag-of-words paradigm could benefit from using this annotation to draw upon (the millions of) Wikipedia pages and their inter-relations.",
"Interlinking text documents with Linked Open Data enables the Web of Data to be used as background knowledge within document-oriented applications such as search and faceted browsing. As a step towards interconnecting the Web of Documents with the Web of Data, we developed DBpedia Spotlight, a system for automatically annotating text documents with DBpedia URIs. DBpedia Spotlight allows users to configure the annotations to their specific needs through the DBpedia Ontology and quality measures such as prominence, topical pertinence, contextual ambiguity and disambiguation confidence. We compare our approach with the state of the art in disambiguation, and evaluate our results in light of three baselines and six publicly available annotation systems, demonstrating the competitiveness of our system. DBpedia Spotlight is shared as open source and deployed as a Web Service freely available for public use.",
"Information extraction from microblog posts is an important task, as today microblogs capture an unprecedented amount of information and provide a view into the pulse of the world. As the core component of information extraction, we consider the task of Twitter entity linking in this paper. In the current entity linking literature, mention detection and entity disambiguation are frequently cast as equally important but distinct problems. However, in our task, we find that mention detection is often the performance bottleneck. The reason is that messages on micro-blogs are short, noisy and informal texts with little context, and often contain phrases with ambiguous meanings. To rigorously address the Twitter entity linking problem, we propose a structural SVM algorithm for entity linking that jointly optimizes mention detection and entity disambiguation as a single end-to-end task. By combining structural learning and a variety of firstorder, second-order, and context-sensitive features, our system is able to outperform existing state-of-the art entity linking systems by 15 F1.",
"Applying natural language processing for mining and intelligent information access to tweets (a form of microblog) is a challenging, emerging research area. Unlike carefully authored news text and other longer content, tweets pose a number of new challenges, due to their short, noisy, context-dependent, and dynamic nature. Information extraction from tweets is typically performed in a pipeline, comprising consecutive stages of language identification, tokenisation, part-of-speech tagging, named entity recognition and entity disambiguation (e.g. with respect to DBpedia). In this work, we describe a new Twitter entity disambiguation dataset, and conduct an empirical analysis of named entity recognition and disambiguation, investigating how robust a number of state-of-the-art systems are on such noisy texts, what the main sources of error are, and which problems should be further investigated to improve the state of the art.",
"We study the task of entity linking for tweets, which tries to associate each mention in a tweet with a knowledge base entry. Two main challenges of this task are the dearth of information in a single tweet and the rich entity mention variations. To address these challenges, we propose a collective inference method that simultaneously resolves a set of mentions. Particularly, our model integrates three kinds of similarities, i.e., mention-entry similarity, entry-entry similarity, and mention-mention similarity, to enrich the context for entity linking, and to address irregular mentions that are not covered by the entity-variation dictionary. We evaluate our method on a publicly available data set and demonstrate the effectiveness of our method.",
"This paper presents a large-scale system for the recognition and semantic disambiguation of named entities based on information extracted from a large encyclopedic collection and Web search results. It describes in detail the disambiguation paradigm employed and the information extraction process from Wikipedia. Through a process of maximizing the agreement between the contextual information extracted from Wikipedia and the context of a document, as well as the agreement among the category tags associated with the candidate entities, the implemented system shows high disambiguation accuracy on both news stories and Wikipedia articles.",
"The integration of facts derived from information extraction systems into existing knowledge bases requires a system to disambiguate entity mentions in the text. This is challenging due to issues such as non-uniform variations in entity names, mention ambiguity, and entities absent from a knowledge base. We present a state of the art system for entity disambiguation that not only addresses these challenges but also scales to knowledge bases with several million entries using very little resources. Further, our approach achieves performance of up to 95 on entities mentioned from newswire and 80 on a public test set that was designed to include challenging queries.",
"We present E2E, an end-to-end entity linking system that is designed for short and noisy text found in microblogs and text messages. Mining and extracting entities from short text is an essential step for many content analysis applications. By jointly optimizing entity recognition and disambiguation as a single task, our system can process short and noisy text robustly."
]
} |
1707.08470 | 2460515658 | Over the years, Twitter has become one of the largest communication platforms providing key data to various applications such as brand monitoring, trend detection, among others. Entity linking is one of the major tasks in natural language understanding from tweets and it associates entity mentions in text to corresponding entries in knowledge bases in order to provide unambiguous interpretation and additional context. State-of-the-art techniques have focused on linking explicitly mentioned entities in tweets with reasonable success. However, we argue that in addition to explicit mentions --- i.e. 'The movie Gravity was more expensive than the mars orbiter mission' --- entities movie Gravity can also be mentioned implicitly --- i.e. 'This new space movie is crazy. you must watch it!.' This paper introduces the problem of implicit entity linking in tweets. We propose an approach that models the entities by exploiting their factual and contextual knowledge. We demonstrate how to use these models to perform implicit entity linking on a ground truth dataset with 397 tweets from two domains, namely, Movie and Book. Specifically, we show: 1 the importance of linking implicit entities and its value addition to the standard entity linking task, and 2 the importance of exploiting contextual knowledge associated with an entity for linking their implicit mentions. We also make the ground truth dataset publicly available to foster the research in this new research areacity. | Our previous work on implicit entity linking has dealt with clinical entities in electronic medical records @cite_14 . The main challenge in the medical setting resides in the heterogeneous usage of language by individuals mentioning entities implicitly including their negated mentions. Our approach focused on modeling the entities by using their static definitions and exploiting WordNet as the knowledge base to account for heterogeneity in language usage. The task of implicit entity linking in tweets is different from that of clinical text since, in addition to the differences in nature of the text, the heterogeneity arise with the usage of different characteristics of entities and their association with dynamic context as discussed in sec:understanding . Hence, the static descriptions definitions of entities fall short in linking implicit entities as we will demonstrate in our evaluation. | {
"cite_N": [
"@cite_14"
],
"mid": [
"2169353103"
],
"abstract": [
"With the increasing automation of health care information processing, it has become crucial to extract meaningful information from textual notes in electronic medical records. One of the key challenges is to extract and normalize entity mentions. State-of-the-art approaches have focused on the recognition of entities that are explicitly mentioned in a sentence. However, clinical documents often contain phrases that indicate the entities but do not contain their names. We term those implicit entity mentions and introduce the problem of implicit entity recognition (IER) in clinical documents. We propose a solution to IER that leverages entity definitions from a knowledge base to create entity models, projects sentences to the entity models and identifies implicit entity mentions by evaluating semantic similarity between sentences and entity models. The evaluation with 857 sentences selected for 8 different entities shows that our algorithm outperforms the most closely related unsupervised solution. The similarity value calculated by our algorithm proved to be an effective feature in a supervised learning setting, helping it to improve over the baselines, and achieving F1 scores of .81 and .73 for different classes of implicit mentions. Our gold standard annotations are made available to encourage further research in the area of IER."
]
} |
1707.08270 | 2741617588 | The concept of bounded highway dimension was developed to capture observed properties of the metrics of road networks. We show that a graph with bounded highway dimension, for any vertex, can be embedded into a a graph of bounded treewidth in such a way that the distance between @math and @math is preserved up to an additive error of @math times the distance from @math or @math to the selected vertex. We show that this theorem yields a PTAS for Bounded-Capacity Vehicle Routing in graphs of bounded highway dimension. In this problem, the input specifies a depot and a set of clients, each with a location and demand; the output is a set of depot-to-depot tours, where each client is visited by some tour and each tour covers at most @math units of client demand. Our PTAS can be extended to handle penalties for unvisited clients. We extend this embedding result to handle a set @math of distinguished vertices. The treewidth depends on @math , and the distance between @math and @math is preserved up to an additive error of @math times the distance from @math and @math to @math . This embedding result implies a PTAS for Multiple Depot Bounded-Capacity Vehicle Routing: the tours can go from one depot to another. The embedding result also implies that, for fixed @math , there is a PTAS for @math -Center in graphs of bounded highway dimension. In this problem, the goal is to minimize @math such that there exist @math vertices (the centers) such that every vertex is within distance @math of some center. Similarly, for fixed @math , there is a PTAS for @math -Median in graphs of bounded highway dimension. In this problem, the goal is to minimize the sum of distances to the @math centers. | * Vehicle routing Haimovich and Rinnoy Kan @cite_33 proved the following lower bound Although their result addresses the unit-demand case, it generalizes to instances where all non-zero client demand @math is at least one for every client @math . : | {
"cite_N": [
"@cite_33"
],
"mid": [
"2155735286"
],
"abstract": [
"In a capacitated routing problem, the objective is to minimize the total distance travelled by vehicles of limited capacity to serve a set of customers that are located in the Euclidean plane. We develop asymptotically optimal bounds and heuristics for this problem, under the assumption that the capacity of a vehicle is expressed in terms of an upper bound on the number of customers that it can serve. The analysis culminates in an algorithm that, for a given capacity and given (epsilon), will find a solution with relative error at most (epsilon) in time polynomial in the number of customers."
]
} |
1707.08270 | 2741617588 | The concept of bounded highway dimension was developed to capture observed properties of the metrics of road networks. We show that a graph with bounded highway dimension, for any vertex, can be embedded into a a graph of bounded treewidth in such a way that the distance between @math and @math is preserved up to an additive error of @math times the distance from @math or @math to the selected vertex. We show that this theorem yields a PTAS for Bounded-Capacity Vehicle Routing in graphs of bounded highway dimension. In this problem, the input specifies a depot and a set of clients, each with a location and demand; the output is a set of depot-to-depot tours, where each client is visited by some tour and each tour covers at most @math units of client demand. Our PTAS can be extended to handle penalties for unvisited clients. We extend this embedding result to handle a set @math of distinguished vertices. The treewidth depends on @math , and the distance between @math and @math is preserved up to an additive error of @math times the distance from @math and @math to @math . This embedding result implies a PTAS for Multiple Depot Bounded-Capacity Vehicle Routing: the tours can go from one depot to another. The embedding result also implies that, for fixed @math , there is a PTAS for @math -Center in graphs of bounded highway dimension. In this problem, the goal is to minimize @math such that there exist @math vertices (the centers) such that every vertex is within distance @math of some center. Similarly, for fixed @math , there is a PTAS for @math -Median in graphs of bounded highway dimension. In this problem, the goal is to minimize the sum of distances to the @math centers. | Little is known for higher dimensions or other metrics. Kachay gave a PTAS in @math that requires @math to be @math @cite_13 , and Hamaguchi and Katoh @cite_16 and Asano, Katoh, and Kawashima @cite_25 focused on constant-factor approximation algorithms for the case where the graph is a tree and client demand is divisible. Becker, Klein and Saulpic @cite_32 gave the first approximation scheme for a non-Euclidean metric: they describe a quasi-polynomial-time approximation scheme in planar graphs, but only when the capacity @math is polylogarithmic in the graph size. They introduce the idea of an error that depends on the distance to the depot, which we also use in the embedding presented in our work here. | {
"cite_N": [
"@cite_32",
"@cite_16",
"@cite_13",
"@cite_25"
],
"mid": [
"2966058519",
"2137208273",
"2559541197",
"1529079830"
],
"abstract": [
"",
"In this paper we deal with a vehicle routing problem on a tree-shaped network with a single depot. Customers are located on vertices of the tree, and each customer has a positive demand. Demands of customers are served by a fleet of identical vehicles with limited capacity. It is assumed that the demand of a customer is splittable, i.e., it can be served by more than one vehicle. The problem we are concerned with in this paper asks to find a set of tours of the vehicles with minimum total lengths. Each tour begins at the depot, visits a subset of the customers and returns to the depot without violating the capacity constraint. We show that the problem is NP-complete and propose a 1.5-approximation algorithm for the problem. We also give some computational results.",
"Capacitated Vehicle Routing Problem (CVRP) is the well-known combinatorial optimization problem remaining NP-hard even in the Euclidean spaces of fixed dimension. Thirty years ago, in their celebrated paper, M. Haimovich and A. Rinnoy Kan proposed the first PTAS for the Planar Single Depot CVRP based on their Iterated Tour Partition heuristic. For decades, this result was extended by many authors to numerous useful modifications of the problem taking into account multiple depots, pick up and delivery options, time window restrictions, etc. But, to the best of our knowledge, almost none of these results go beyond the Euclidean plane. In this paper, we try to bridge this gap and propose an EPTAS for the Euclidean CVRP for any fixed dimension.",
"This paper presents a new approximation algorithm for a vehicle routing problem on a tree-shaped network with a single depot. Customers are located on vertices of the tree, and each customer has a positive demand. Demands of customers are served by a fleet of identical vehicles with limited capacity. It is assumed that the demand of a customer is splittable, i.e., it can be served by more than one vehicle. The problem we are concerned with in this paper asks to find a set of tours of the vehicles with minimum total lengths. Each tour begins at the depot, visits a subset of the customers and returns to the depot without violating the capacity constraint We propose a 1.35078-approximation algorithm for the problem (exactly, (√41 - 1) 4), which is an improvement over the existing 1.5-approximation."
]
} |
1707.08270 | 2741617588 | The concept of bounded highway dimension was developed to capture observed properties of the metrics of road networks. We show that a graph with bounded highway dimension, for any vertex, can be embedded into a a graph of bounded treewidth in such a way that the distance between @math and @math is preserved up to an additive error of @math times the distance from @math or @math to the selected vertex. We show that this theorem yields a PTAS for Bounded-Capacity Vehicle Routing in graphs of bounded highway dimension. In this problem, the input specifies a depot and a set of clients, each with a location and demand; the output is a set of depot-to-depot tours, where each client is visited by some tour and each tour covers at most @math units of client demand. Our PTAS can be extended to handle penalties for unvisited clients. We extend this embedding result to handle a set @math of distinguished vertices. The treewidth depends on @math , and the distance between @math and @math is preserved up to an additive error of @math times the distance from @math and @math to @math . This embedding result implies a PTAS for Multiple Depot Bounded-Capacity Vehicle Routing: the tours can go from one depot to another. The embedding result also implies that, for fixed @math , there is a PTAS for @math -Center in graphs of bounded highway dimension. In this problem, the goal is to minimize @math such that there exist @math vertices (the centers) such that every vertex is within distance @math of some center. Similarly, for fixed @math , there is a PTAS for @math -Median in graphs of bounded highway dimension. In this problem, the goal is to minimize the sum of distances to the @math centers. | * @math - Median We have already surveyed some of the results on @math - Center . Note that the results of Feldmann @cite_9 are based on the definition of highway dimension of 2011 @cite_35 , but can be adapted to the definition we use here. | {
"cite_N": [
"@cite_35",
"@cite_9"
],
"mid": [
"85521454",
"2949966487"
],
"abstract": [
"We explore the relationship between VC-dimension and graph algorithm design. In particular, we show that set systems induced by sets of vertices on shortest paths have VC-dimension at most two. This allows us to use a result from learning theory to improve time bounds on query algorithms for the point-to-point shortest path problem in networks of low highway dimension, such as road networks. We also refine the definitions of highway dimension and related concepts, making them more general and potentially more relevant to practice. In particular, we define highway dimension in terms of set systems induced by shortest paths, and give cardinality-based and average case definitions.",
"We consider the @math -Center problem and some generalizations. For @math -Center a set of @math center vertices needs to be found in a graph @math with edge lengths, such that the distance from any vertex of @math to its nearest center is minimized. This problem naturally occurs in transportation networks, and therefore we model the inputs as graphs with bounded highway dimension, as proposed by [SODA 2010]. We show both approximation and fixed-parameter hardness results, and how to overcome them using fixed-parameter approximations, where the two paradigms are combined. In particular, we prove that for any @math computing a @math -approximation is W[2]-hard for parameter @math and NP-hard for graphs with highway dimension @math . The latter does not rule out fixed-parameter @math -approximations for the highway dimension parameter @math , but implies that such an algorithm must have at least doubly exponential running time in @math if it exists, unless the ETH fails. On the positive side, we show how to get below the approximation factor of @math by combining the parameters @math and @math : we develop a fixed-parameter @math -approximation with running time @math . Additionally we prove that, unless P=NP, our techniques cannot be used to compute fixed-parameter @math -approximations for only the parameter @math . We also provide similar fixed-parameter approximations for the weighted @math -Center and @math -Partition problems, which generalize @math -Center."
]
} |
1707.08270 | 2741617588 | The concept of bounded highway dimension was developed to capture observed properties of the metrics of road networks. We show that a graph with bounded highway dimension, for any vertex, can be embedded into a a graph of bounded treewidth in such a way that the distance between @math and @math is preserved up to an additive error of @math times the distance from @math or @math to the selected vertex. We show that this theorem yields a PTAS for Bounded-Capacity Vehicle Routing in graphs of bounded highway dimension. In this problem, the input specifies a depot and a set of clients, each with a location and demand; the output is a set of depot-to-depot tours, where each client is visited by some tour and each tour covers at most @math units of client demand. Our PTAS can be extended to handle penalties for unvisited clients. We extend this embedding result to handle a set @math of distinguished vertices. The treewidth depends on @math , and the distance between @math and @math is preserved up to an additive error of @math times the distance from @math and @math to @math . This embedding result implies a PTAS for Multiple Depot Bounded-Capacity Vehicle Routing: the tours can go from one depot to another. The embedding result also implies that, for fixed @math , there is a PTAS for @math -Center in graphs of bounded highway dimension. In this problem, the goal is to minimize @math such that there exist @math vertices (the centers) such that every vertex is within distance @math of some center. Similarly, for fixed @math , there is a PTAS for @math -Median in graphs of bounded highway dimension. In this problem, the goal is to minimize the sum of distances to the @math centers. | For @math -Median , constant-factor approximation algorithms have been found for general metric spaces @cite_11 @cite_34 @cite_7 . The best known approximation ratio for @math -Median in general metrics is @math @cite_19 , and it is NP-hard to approximate within a factor of @math @cite_20 . For @math -Median in @math -dimensional Euclidean space, PTAS have been found when @math is fixed (e.g. @cite_24 ) and when @math is fixed (e.g. @cite_23 ) but there exists no PTAS if @math and @math are part of the input @cite_14 . Recently Cohen- @cite_29 gave a local search-based PTAS for @math -Median in edge-weight planar graphs that extends more generally to minor-closed graph families. | {
"cite_N": [
"@cite_14",
"@cite_7",
"@cite_29",
"@cite_24",
"@cite_19",
"@cite_23",
"@cite_34",
"@cite_20",
"@cite_11"
],
"mid": [
"2010480966",
"",
"2962937493",
"2059651397",
"2343942810",
"1977541023",
"",
"2101622070",
"2147670675"
],
"abstract": [
"",
"",
"We give the first polynomial-time approximation schemes (PTASs) for the following problems: (1) uniform facility location in edge-weighted planar graphs, (2) k-median and k-means in edge-weighted planar graphs, (3) k-means in Euclidean space of bounded dimension. Our first and second results extend to minor-closed families of graphs. All our results extend to cost functions that are the pth power of the shortest-path distance. The algorithm is local search where the local neighborhood of a solution S consists of all solutions obtained from S by removing and adding 1 eO(1) centers.",
"In this paper, we show that for several clustering problems one can extract a small set of points, so that using those core-sets enable us to perform approximate clustering efficiently. The surprising property of those core-sets is that their size is independent of the dimension.Using those, we present a (1+ e)-approximation algorithms for the k-center clustering and k-median clustering problems in Euclidean space. The running time of the new algorithms has linear or near linear dependency on the number of points and the dimension, and exponential dependency on 1 e and k. As such, our results are a substantial improvement over what was previously known.We also present some other clustering results including (1+ e)-approximate 1-cylinder clustering, and k-center clustering with outliers.",
"We present a novel approximation algorithm for @math -median that achieves an approximation guarantee of @math , improving upon the decade-old ratio of @math . Our improved approximation ratio is achieved by exploiting the power of pseudo-approximation. More specifically, our approach is based on two components, each of which, we believe, is of independent interest. First, we show that in order to give an @math -approximation algorithm for @math -median, it is sufficient to give a pseudo-approximation algorithm that finds an @math -approximate solution by opening @math facilities. This is a rather surprising result as there exist instances for which opening @math facilities may lead to a significantly smaller cost than that of opening only @math facilities. Second, we give such a pseudo-approximation algorithm with @math . Prior to our work, it was not even known whether opening @math facilities would help improve the approximation ratio.",
"In the k-median problem we are given a set S of n points in a metric space and a positive integer k. We desire to locate k medians in space, such that the sum of the distances from each of the points of S to the nearest median is minimized. This paper gives an approximation scheme for the plane that for any c > 0 produces a solution of cost at most 1+ 1 c times the optimum and runs in time O(n). The approximation scheme also generalizes to some problems related to k-median. Our methodology is to extend Arora’s [1, 2] techniques for the TSP, which hitherto seemed inapplicable to problems such as the k-median problem.",
"",
"A fundamental facility location problem is to choose the location of facilities, such as industrial plants and warehouses, to minimize the cost of satisfying the demand for some commodity. There are associated costs for locating the facilities, as well as transportation costs for distributing the commodities. We assume that the transportation costs form a metric. This problem is commonly referred to as theuncapacitated facility locationproblem. Application to bank account location and clustering, as well as many related pieces of work, are discussed by Cornuejols, Nemhauser, and Wolsey. Recently, the first constant factor approximation algorithm for this problem was obtained by Shmoys, Tardos, and Aardal. We show that a simple greedy heuristic combined with the algorithm by Shmoys, Tardos, and Aardal, can be used to obtain an approximation guarantee of 2.408. We discuss a few variants of the problem, demonstrating better approximation factors for restricted versions of the problem. We also show that the problem is max SNP-hard. However, the inapproximability constants derived from the max SNP hardness are very close to one. By relating this problem to Set Cover, we prove a lower bound of 1.463 on the best possible approximation ratio, assumingNP?DTIMEnO(loglogn)].",
"One of the most flourishing areas of research in the design and analysis of approximation algorithms has been for facility location problems. In particular, for the metric case of two simple models, the uncapacitated facility location and the k-median problems, there are now a variety of techniques that yield constant performance guarantees. These methods include LP rounding, primal-dual algorithms, and local search techniques. Furthermore, the salient ideas in these algorithms and their analyzes are simple-to-explain and reflect a surprising degree of commonality. This note is intended as companion to our lecture at CONF 2000, mainly to give pointers to the appropriate references."
]
} |
1707.08015 | 2603292144 | Each year, thousands of software vulnerabilities are discovered and reported to the public. Unpatched known vulnerabilities are a significant security risk. It is imperative that software vendors quickly provide patches once vulnerabilities are known and users quickly install those patches as soon as they are available. However, most vulnerabilities are never actually exploited. Since writing, testing, and installing software patches can involve considerable resources, it would be desirable to prioritize the remediation of vulnerabilities that are likely to be exploited. Several published research studies have reported moderate success in applying machine learning techniques to the task of predicting whether a vulnerability will be exploited. These approaches typically use features derived from vulnerability databases (such as the summary text describing the vulnerability) or social media posts that mention the vulnerability by name. However, these prior studies share multiple methodological shortcomings that inflate predictive power of these approaches. We replicate key portions of the prior work, compare their approaches, and show how selection of training and test data critically affect the estimated performance of predictive models. The results of this study point to important methodological considerations that should be taken into account so that results reflect real-world utility. | Much of the prior work in exploit prediction takes an approach at odds with one or more the aforementioned principles. Bozorgi @cite_0 and Edkrantz @cite_7 each evaluate their methods using artificially balanced datasets, inflating performance metrics. Edkrantz @cite_7 and Sabottke @cite_15 each randomly split training and test data, allowing the temporal intermixing of future'' and past'' data in the training and test sets. Many of the most influential features in the model used by Bozorgi @cite_0 are derived from information that would not be available in an online setting (though they do present results from a simulated online experiment). Sabottke @cite_15 evaluates the utility of text-based features derived from social media, but never compares those results to using text features derived from the vulnerability summary that is readily available at the time of disclosure. | {
"cite_N": [
"@cite_0",
"@cite_15",
"@cite_7"
],
"mid": [
"1971733255",
"1707806712",
"2402739313"
],
"abstract": [
"The security demands on modern system administration are enormous and getting worse. Chief among these demands, administrators must monitor the continual ongoing disclosure of software vulnerabilities that have the potential to compromise their systems in some way. Such vulnerabilities include buffer overflow errors, improperly validated inputs, and other unanticipated attack modalities. In 2008, over 7,400 new vulnerabilities were disclosed--well over 100 per week. While no enterprise is affected by all of these disclosures, administrators commonly face many outstanding vulnerabilities across the software systems they manage. Vulnerabilities can be addressed by patches, reconfigurations, and other workarounds; however, these actions may incur down-time or unforeseen side-effects. Thus, a key question for systems administrators is which vulnerabilities to prioritize. From publicly available databases that document past vulnerabilities, we show how to train classifiers that predict whether and how soon a vulnerability is likely to be exploited. As input, our classifiers operate on high dimensional feature vectors that we extract from the text fields, time stamps, cross references, and other entries in existing vulnerability disclosure reports. Compared to current industry-standard heuristics based on expert knowledge and static formulas, our classifiers predict much more accurately whether and how soon individual vulnerabilities are likely to be exploited.",
"In recent years, the number of software vulnerabilities discovered has grown significantly. This creates a need for prioritizing the response to new disclosures by assessing which vulnerabilities are likely to be exploited and by quickly ruling out the vulnerabilities that are not actually exploited in the real world. We conduct a quantitative and qualitative exploration of the vulnerability-related information disseminated on Twitter. We then describe the design of a Twitter-based exploit detector, and we introduce a threat model specific to our problem. In addition to response prioritization, our detection techniques have applications in risk modeling for cyber-insurance and they highlight the value of information provided by the victims of attacks.",
""
]
} |
1707.07901 | 2738463471 | Adversarial learning has been successfully embedded into deep networks to learn transferable features, which reduce distribution discrepancy between the source and target domains. Existing domain adversarial networks assume fully shared label space across domains. In the presence of big data, there is strong motivation of transferring both classification and representation models from existing big domains to unknown small domains. This paper introduces partial transfer learning, which relaxes the shared label space assumption to that the target label space is only a subspace of the source label space. Previous methods typically match the whole source domain to the target domain, which are prone to negative transfer for the partial transfer problem. We present Selective Adversarial Network (SAN), which simultaneously circumvents negative transfer by selecting out the outlier source classes and promotes positive transfer by maximally matching the data distributions in the shared label space. Experiments demonstrate that our models exceed state-of-the-art results for partial transfer learning tasks on several benchmark datasets. | Transfer learning @cite_13 bridges different domains or tasks to mitigate the burden of manual labeling for machine learning @cite_0 @cite_6 @cite_25 @cite_4 , computer vision @cite_7 @cite_8 @cite_23 and natural language processing @cite_12 . The main technical difficulty of transfer learning is to formally reduce the distribution discrepancy across different domains. Deep networks can learn abstract representations that disentangle different explanatory factors of variations behind data @cite_24 and manifest invariant factors underlying different populations that transfer well from original tasks to similar novel tasks @cite_28 . Thus deep networks have been explored for transfer learning @cite_21 @cite_1 @cite_23 , multimodal and multi-task learning @cite_12 @cite_11 , where significant performance gains have been witnessed relative to prior shallow transfer learning methods. | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_28",
"@cite_21",
"@cite_1",
"@cite_6",
"@cite_0",
"@cite_24",
"@cite_23",
"@cite_13",
"@cite_25",
"@cite_12",
"@cite_11"
],
"mid": [
"2147520416",
"1722318740",
"",
"2949667497",
"22861983",
"2161381512",
"2120149881",
"2115403315",
"2163922914",
"2951084305",
"2165698076",
"2153929442",
"2158899491",
"2184188583"
],
"abstract": [
"Transfer learning algorithms are used when one has sufficient training data for one supervised learning task (the source training domain) but only very limited training data for a second task (the target test domain) that is similar but not identical to the first. Previous work on transfer learning has focused on relatively restricted settings, where specific parts of the model are considered to be carried over between tasks. Recent work on covariate shift focuses on matching the marginal distributions on observations X across domains. Similarly, work on target conditional shift focuses on matching marginal distributions on labels Y and adjusting conditional distributions P(X|Y ), such that P(X) can be matched across domains. However, covariate shift assumes that the support of test P(X) is contained in the support of training P(X), i.e., the training set is richer than the test set. Target conditional shift makes a similar assumption for P(Y). Moreover, not much work on transfer learning has considered the case when a few labels in the test domain are available. Also little work has been done when all marginal and conditional distributions are allowed to change while the changes are smooth. In this paper, we consider a general case where both the support and the model change across domains. We transform both X and Y by a location-scale shift to achieve transfer between tasks. Since we allow more flexible transformations, the proposed method yields better results on both synthetic data and real-world data.",
"Domain adaptation is an important emerging topic in computer vision. In this paper, we present one of the first studies of domain shift in the context of object recognition. We introduce a method that adapts object models acquired in a particular visual domain to new imaging conditions by learning a transformation that minimizes the effect of domain-induced changes in the feature distribution. The transformation is learned in a supervised manner and can be applied to categories for which there are no labeled examples in the new domain. While we focus our evaluation on object recognition tasks, the transform-based adaptation technique we develop is general and could be applied to nonimage data. Another contribution is a new multi-domain object database, freely available for download. We experimentally demonstrate the ability of our method to improve recognition on categories with few or no target domain labels and moderate to large changes in the imaging conditions.",
"",
"Many deep neural networks trained on natural images exhibit a curious phenomenon in common: on the first layer they learn features similar to Gabor filters and color blobs. Such first-layer features appear not to be specific to a particular dataset or task, but general in that they are applicable to many datasets and tasks. Features must eventually transition from general to specific by the last layer of the network, but this transition has not been studied extensively. In this paper we experimentally quantify the generality versus specificity of neurons in each layer of a deep convolutional neural network and report a few surprising results. Transferability is negatively affected by two distinct issues: (1) the specialization of higher layer neurons to their original task at the expense of performance on the target task, which was expected, and (2) optimization difficulties related to splitting networks between co-adapted neurons, which was not expected. In an example network trained on ImageNet, we demonstrate that either of these two issues may dominate, depending on whether features are transferred from the bottom, middle, or top of the network. We also document that the transferability of features decreases as the distance between the base task and target task increases, but that transferring features even from distant tasks can be better than using random features. A final surprising result is that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset.",
"The exponential increase in the availability of online reviews and recommendations makes sentiment classification an interesting topic in academic and industrial research. Reviews can span so many different domains that it is difficult to gather annotated training data for all of them. Hence, this paper studies the problem of domain adaptation for sentiment classifiers, hereby a system is trained on labeled reviews from one source domain but is meant to be deployed on another. We propose a deep learning approach which learns to extract a meaningful representation for each review in an unsupervised fashion. Sentiment classifiers trained with this high-level feature representation clearly outperform state-of-the-art methods on a benchmark composed of reviews of 4 types of Amazon products. Furthermore, this method scales well and allowed us to successfully perform domain adaptation on a larger industrial-strength dataset of 22 domains.",
"Convolutional neural networks (CNN) have recently shown outstanding image classification performance in the large- scale visual recognition challenge (ILSVRC2012). The suc- cess of CNNs is attributed to their ability to learn rich mid- level image representations as opposed to hand-designed low-level features used in other image classification meth- ods. Learning CNNs, however, amounts to estimating mil- lions of parameters and requires a very large number of annotated image samples. This property currently prevents application of CNNs to problems with limited training data. In this work we show how image representations learned with CNNs on large-scale annotated datasets can be effi- ciently transferred to other visual recognition tasks with limited amount of training data. We design a method to reuse layers trained on the ImageNet dataset to compute mid-level image representation for images in the PASCAL VOC dataset. We show that despite differences in image statistics and tasks in the two datasets, the transferred rep- resentation leads to significantly improved results for object and action classification, outperforming the current state of the art on Pascal VOC 2007 and 2012 datasets. We also show promising results for object and action localization.",
"Cross-domain learning methods have shown promising results by leveraging labeled patterns from the auxiliary domain to learn a robust classifier for the target domain which has only a limited number of labeled samples. To cope with the considerable change between feature distributions of different domains, we propose a new cross-domain kernel learning framework into which many existing kernel methods can be readily incorporated. Our framework, referred to as Domain Transfer Multiple Kernel Learning (DTMKL), simultaneously learns a kernel function and a robust classifier by minimizing both the structural risk functional and the distribution mismatch between the labeled and unlabeled samples from the auxiliary and target domains. Under the DTMKL framework, we also propose two novel methods by using SVM and prelearned classifiers, respectively. Comprehensive experiments on three domain adaptation data sets (i.e., TRECVID, 20 Newsgroups, and email spam data sets) demonstrate that DTMKL-based methods outperform existing cross-domain learning and multiple kernel learning methods.",
"Domain adaptation allows knowledge from a source domain to be transferred to a different but related target domain. Intuitively, discovering a good feature representation across domains is crucial. In this paper, we first propose to find such a representation through a new learning method, transfer component analysis (TCA), for domain adaptation. TCA tries to learn some transfer components across domains in a reproducing kernel Hilbert space using maximum mean miscrepancy. In the subspace spanned by these transfer components, data properties are preserved and data distributions in different domains are close to each other. As a result, with the new representations in this subspace, we can apply standard machine learning methods to train classifiers or regression models in the source domain for use in the target domain. Furthermore, in order to uncover the knowledge hidden in the relations between the data labels from the source and target domains, we extend TCA in a semisupervised learning setting, which encodes label information into transfer components learning. We call this extension semisupervised TCA. The main contribution of our work is that we propose a novel dimensionality reduction framework for reducing the distance between domains in a latent space for domain adaptation. We propose both unsupervised and semisupervised feature extraction approaches, which can dramatically reduce the distance between domain distributions by projecting data onto the learned transfer components. Finally, our approach can handle large datasets and naturally lead to out-of-sample generalization. The effectiveness and efficiency of our approach are verified by experiments on five toy datasets and two real-world applications: cross-domain indoor WiFi localization and cross-domain text classification.",
"The success of machine learning algorithms generally depends on data representation, and we hypothesize that this is because different representations can entangle and hide more or less the different explanatory factors of variation behind the data. Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning algorithms implementing such priors. This paper reviews recent work in the area of unsupervised feature learning and deep learning, covering advances in probabilistic models, autoencoders, manifold learning, and deep networks. This motivates longer term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between representation learning, density estimation, and manifold learning.",
"A major challenge in scaling object detection is the difficulty of obtaining labeled images for large numbers of categories. Recently, deep convolutional neural networks (CNNs) have emerged as clear winners on object classification benchmarks, in part due to training with 1.2M+ labeled classification images. Unfortunately, only a small fraction of those labels are available for the detection task. It is much cheaper and easier to collect large quantities of image-level labels from search engines than it is to collect detection data and label it with precise bounding boxes. In this paper, we propose Large Scale Detection through Adaptation (LSDA), an algorithm which learns the difference between the two tasks and transfers this knowledge to classifiers for categories without bounding box annotated data, turning them into detectors. Our method has the potential to enable detection for the tens of thousands of categories that lack bounding box annotations, yet have plenty of classification data. Evaluation on the ImageNet LSVRC-2013 detection challenge demonstrates the efficacy of our approach. This algorithm enables us to produce a >7.6K detector by using available classification data from leaf nodes in the ImageNet tree. We additionally demonstrate how to modify our architecture to produce a fast detector (running at 2fps for the 7.6K detector). Models and software are available at",
"A major assumption in many machine learning and data mining algorithms is that the training and future data must be in the same feature space and have the same distribution. However, in many real-world applications, this assumption may not hold. For example, we sometimes have a classification task in one domain of interest, but we only have sufficient training data in another domain of interest, where the latter data may be in a different feature space or follow a different data distribution. In such cases, knowledge transfer, if done successfully, would greatly improve the performance of learning by avoiding much expensive data-labeling efforts. In recent years, transfer learning has emerged as a new learning framework to address this problem. This survey focuses on categorizing and reviewing the current progress on transfer learning for classification, regression, and clustering problems. In this survey, we discuss the relationship between transfer learning and other related machine learning techniques such as domain adaptation, multitask learning and sample selection bias, as well as covariate shift. We also explore some potential future issues in transfer learning research.",
"Let X denote the feature and Y the target. We consider domain adaptation under three possible scenarios: (1) the marginal PY changes, while the conditional PX Y stays the same (target shift), (2) the marginal PY is fixed, while the conditional PX Y changes with certain constraints (conditional shift), and (3) the marginal PY changes, and the conditional PX Y changes with constraints (generalized target shift). Using background knowledge, causal interpretations allow us to determine the correct situation for a problem at hand. We exploit importance reweighting or sample transformation to find the learning machine that works well on test data, and propose to estimate the weights or transformations by reweighting or transforming training data to reproduce the covariate distribution on the test domain. Thanks to kernel embedding of conditional as well as marginal distributions, the proposed approaches avoid distribution estimation, and are applicable for high-dimensional problems. Numerical evaluations on synthetic and real-world data sets demonstrate the effectiveness of the proposed framework.",
"We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements.",
"Deep networks have been successfully applied to unsupervised feature learning for single modalities (e.g., text, images or audio). In this work, we propose a novel application of deep networks to learn features over multiple modalities. We present a series of tasks for multimodal learning and show how to train deep networks that learn features to address these tasks. In particular, we demonstrate cross modality feature learning, where better features for one modality (e.g., video) can be learned if multiple modalities (e.g., audio and video) are present at feature learning time. Furthermore, we show how to learn a shared representation between modalities and evaluate it on a unique task, where the classifier is trained with audio-only data but tested with video-only data and vice-versa. Our models are validated on the CUAVE and AVLetters datasets on audio-visual speech classification, demonstrating best published visual speech classification on AVLetters and effective shared representation learning."
]
} |
1707.07901 | 2738463471 | Adversarial learning has been successfully embedded into deep networks to learn transferable features, which reduce distribution discrepancy between the source and target domains. Existing domain adversarial networks assume fully shared label space across domains. In the presence of big data, there is strong motivation of transferring both classification and representation models from existing big domains to unknown small domains. This paper introduces partial transfer learning, which relaxes the shared label space assumption to that the target label space is only a subspace of the source label space. Previous methods typically match the whole source domain to the target domain, which are prone to negative transfer for the partial transfer problem. We present Selective Adversarial Network (SAN), which simultaneously circumvents negative transfer by selecting out the outlier source classes and promotes positive transfer by maximally matching the data distributions in the shared label space. Experiments demonstrate that our models exceed state-of-the-art results for partial transfer learning tasks on several benchmark datasets. | However, recent advances show that deep networks can learn abstract feature representations that can only reduce, but not remove, the cross-domain discrepancy @cite_21 @cite_26 , resulting in unbounded risk for target tasks @cite_2 @cite_16 . Some recent work bridges deep learning and domain adaptation @cite_26 @cite_20 @cite_18 @cite_10 @cite_15 , which extends deep convolutional networks (CNNs) to domain adaptation by adding adaptation layers through which the mean embeddings of distributions are matched @cite_26 @cite_20 @cite_15 , or by adding a subnetwork as domain discriminator while the deep features are learned to confuse the discriminator in a domain-adversarial training paradigm @cite_18 @cite_10 . While performance was significantly improved, these state of the art methods may be restricted by the assumption that the source and target domains share the same label space. This assumption is violated in partial transfer learning, which transfers both representation and classification models from existing big domains to unknown small domains. To our knowledge, this is the first work that addresses partial transfer learning in adversarial networks. | {
"cite_N": [
"@cite_18",
"@cite_26",
"@cite_21",
"@cite_2",
"@cite_15",
"@cite_16",
"@cite_10",
"@cite_20"
],
"mid": [
"1882958252",
"1565327149",
"22861983",
"2953369858",
"2950361018",
"2104094955",
"2953226914",
"2951670162"
],
"abstract": [
"Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of \"deep\" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation. Overall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets.",
"Recent reports suggest that a generic supervised deep CNN model trained on a large-scale dataset reduces, but does not remove, dataset bias on a standard benchmark. Fine-tuning deep models in a new domain can require a significant amount of data, which for many applications is simply not available. We propose a new CNN architecture which introduces an adaptation layer and an additional domain confusion loss, to learn a representation that is both semantically meaningful and domain invariant. We additionally show that a domain confusion metric can be used for model selection to determine the dimension of an adaptation layer and the best position for the layer in the CNN architecture. Our proposed adaptation method offers empirical performance which exceeds previously published results on a standard benchmark visual domain adaptation task.",
"The exponential increase in the availability of online reviews and recommendations makes sentiment classification an interesting topic in academic and industrial research. Reviews can span so many different domains that it is difficult to gather annotated training data for all of them. Hence, this paper studies the problem of domain adaptation for sentiment classifiers, hereby a system is trained on labeled reviews from one source domain but is meant to be deployed on another. We propose a deep learning approach which learns to extract a meaningful representation for each review in an unsupervised fashion. Sentiment classifiers trained with this high-level feature representation clearly outperform state-of-the-art methods on a benchmark composed of reviews of 4 types of Amazon products. Furthermore, this method scales well and allowed us to successfully perform domain adaptation on a larger industrial-strength dataset of 22 domains.",
"This paper addresses the general problem of domain adaptation which arises in a variety of applications where the distribution of the labeled sample available somewhat differs from that of the test data. Building on previous work by Ben- (2007), we introduce a novel distance between distributions, discrepancy distance, that is tailored to adaptation problems with arbitrary loss functions. We give Rademacher complexity bounds for estimating the discrepancy distance from finite samples for different loss functions. Using this distance, we derive novel generalization bounds for domain adaptation for a wide family of loss functions. We also present a series of novel adaptation bounds for large classes of regularization-based algorithms, including support vector machines and kernel ridge regression based on the empirical discrepancy. This motivates our analysis of the problem of minimizing the empirical discrepancy for various loss functions for which we also give novel algorithms. We report the results of preliminary experiments that demonstrate the benefits of our discrepancy minimization algorithms for domain adaptation.",
"The recent success of deep neural networks relies on massive amounts of labeled data. For a target task where labeled data is unavailable, domain adaptation can transfer a learner from a different source domain. In this paper, we propose a new approach to domain adaptation in deep networks that can simultaneously learn adaptive classifiers and transferable features from labeled data in the source domain and unlabeled data in the target domain. We relax a shared-classifier assumption made by previous methods and assume that the source classifier and target classifier differ by a residual function. We enable classifier adaptation by plugging several layers into the deep network to explicitly learn the residual function with reference to the target classifier. We embed features of multiple layers into reproducing kernel Hilbert spaces (RKHSs) and match feature distributions for feature adaptation. The adaptation behaviors can be achieved in most feed-forward models by extending them with new residual layers and loss functions, which can be trained efficiently using standard back-propagation. Empirical evidence exhibits that the approach outperforms state of art methods on standard domain adaptation datasets.",
"Discriminative learning methods for classification perform well when training and test data are drawn from the same distribution. Often, however, we have plentiful labeled training data from a source domain but wish to learn a classifier which performs well on a target domain with a different distribution and little or no labeled training data. In this work we investigate two questions. First, under what conditions can a classifier trained from source data be expected to perform well on target data? Second, given a small amount of labeled target data, how should we combine it during training with the large amount of labeled source data to achieve the lowest target error at test time? We address the first question by bounding a classifier's target error in terms of its source error and the divergence between the two domains. We give a classifier-induced divergence measure that can be estimated from finite, unlabeled samples from the domains. Under the assumption that there exists some hypothesis that performs well in both domains, we show that this quantity together with the empirical source error characterize the target error of a source-trained classifier. We answer the second question by bounding the target error of a model which minimizes a convex combination of the empirical source and target errors. Previous theoretical work has considered minimizing just the source error, just the target error, or weighting instances from the two domains equally. We show how to choose the optimal combination of source and target error as a function of the divergence, the sample sizes of both domains, and the complexity of the hypothesis class. The resulting bound generalizes the previously studied cases and is always at least as tight as a bound which considers minimizing only the target error or an equal weighting of source and target errors.",
"Recent reports suggest that a generic supervised deep CNN model trained on a large-scale dataset reduces, but does not remove, dataset bias. Fine-tuning deep models in a new domain can require a significant amount of labeled data, which for many applications is simply not available. We propose a new CNN architecture to exploit unlabeled and sparsely labeled target domain data. Our approach simultaneously optimizes for domain invariance to facilitate domain transfer and uses a soft label distribution matching loss to transfer information between tasks. Our proposed adaptation method offers empirical performance which exceeds previously published results on two standard benchmark visual domain adaptation tasks, evaluated across supervised and semi-supervised adaptation settings.",
"Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multi-kernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks."
]
} |
1707.08081 | 2736557444 | Search systems are often focused on providing relevant results for the "now", assuming both corpora and user needs that focus on the present. However, many corpora today reflect significant longitudinal collections ranging from 20 years of the Web to hundreds of years of digitized newspapers and books. Understanding the temporal intent of the user and retrieving the most relevant historical content has become a significant challenge. Common search features, such as query expansion, leverage the relationship between terms but cannot function well across all times when relationships vary temporally. In this work, we introduce a temporal relationship model that is extracted from longitudinal data collections. The model supports the task of identifying, given two words, when they relate to each other. We present an algorithmic framework for this task and show its application for the task of query expansion, achieving high gain. | Understanding the semantic change of words has become an active research topic (). Most work has focused on identifying semantic drifts and word meaning changes. A lot of effort has been made into analyzing texts temporally: several methods for temporal information extraction were recently proposed @cite_8 @cite_28 @cite_20 , as well as publicly released knowledge bases, such as YAGO2 @cite_25 . These methods automatically extract temporal relational facts from free text or semi-structured data. In addition, and others annotated texts temporally, and extracted events as well as temporal expressions. | {
"cite_N": [
"@cite_28",
"@cite_25",
"@cite_20",
"@cite_8"
],
"mid": [
"2155069392",
"2122865749",
"2151718942",
""
],
"abstract": [
"Recently, large-scale knowledge bases have been constructed by automatically extracting relational facts from text. Unfortunately, most of the current knowledge bases focus on static facts and ignore the temporal dimension. However, the vast majority of facts are evolving with time or are valid only during a particular time period. Thus, time is a significant dimension that should be included in knowledge bases. In this paper, we introduce a complete information extraction framework that harvests temporal facts and events from semi-structured data and free text of Wikipedia articles to create a temporal ontology. First, we extend a temporal data representation model by making it aware of events. Second, we develop an information extraction method which harvests temporal facts and events from Wikipedia infoboxes, categories, lists, and article titles in order to build a temporal knowledge base. Third, we show how the system can use its extracted knowledge for further growing the knowledge base. We demonstrate the effectiveness of our proposed methods through several experiments. We extracted more than one million temporal facts with precision over 90 for extraction from semi-structured data and almost 70 for extraction from text.",
"We present YAGO2, an extension of the YAGO knowledge base, in which entities, facts, and events are anchored in both time and space. YAGO2 is built automatically from Wikipedia, GeoNames, and WordNet. It contains 447 million facts about 9.8 million entities. Human evaluation confirmed an accuracy of 95 of the facts in YAGO2. In this paper, we present the extraction methodology, the integration of the spatio-temporal dimension, and our knowledge representation SPOTL, an extension of the original SPO-triple model to time and space.",
"Recent research has made significant advances in automatically constructing knowledge bases by extracting relational facts (e.g., Bill Clinton-presidentOf-US) from large text corpora. Temporally scoping such relational facts in the knowledge base (i.e., determining that Bill Clinton-presidentOf-US is true only during the period 1993 - 2001) is an important, but relatively unexplored problem. In this paper, we propose a joint inference framework for this task, which leverages fact-specific temporal constraints, and weak supervision in the form of a few labeled examples. Our proposed framework, CoTS (Coupled Temporal Scoping), exploits temporal containment, alignment, succession, and mutual exclusion constraints among facts from within and across relations. Our contribution is multi-fold. Firstly, while most previous research has focused on micro-reading approaches for temporal scoping, we pose it in a macro-reading fashion, as a change detection in a time series of facts' features computed from a large number of documents. Secondly, to the best of our knowledge, there is no other work that has used joint inference for temporal scoping. We show that joint inference is effective compared to doing temporal scoping of individual facts independently. We conduct our experiments on large scale open-domain publicly available time-stamped datasets, such as English Gigaword Corpus and Google Books Ngrams, demonstrating CoTS's effectiveness.",
""
]
} |
1707.08081 | 2736557444 | Search systems are often focused on providing relevant results for the "now", assuming both corpora and user needs that focus on the present. However, many corpora today reflect significant longitudinal collections ranging from 20 years of the Web to hundreds of years of digitized newspapers and books. Understanding the temporal intent of the user and retrieving the most relevant historical content has become a significant challenge. Common search features, such as query expansion, leverage the relationship between terms but cannot function well across all times when relationships vary temporally. In this work, we introduce a temporal relationship model that is extracted from longitudinal data collections. The model supports the task of identifying, given two words, when they relate to each other. We present an algorithmic framework for this task and show its application for the task of query expansion, achieving high gain. | Continuous word embeddings @cite_18 have been shown to effectively encapsulate relatedness between words. used temporal patterns of words from a large corpus for the task of word similarity. They showed that words that co-occur in history have a stronger relation. In our work, we focus on identifying a relation holds. Numerous projects have studied the change of word meanings over time, and specifically focused on identification of the change itself. used Latent Semantic Analysis for detecting changes in word meaning. characterized 20 clusters to describe the nature of meaning change over time, whereas used other clustering techniques to find changes in word senses. identified changes in word usage over time by the change in their related part-of-speech. Others have investigated the use of word frequency to identify epochs @cite_1 . | {
"cite_N": [
"@cite_18",
"@cite_1"
],
"mid": [
"2950133940",
"2250775745"
],
"abstract": [
"The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling. An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of \"Canada\" and \"Air\" cannot be easily combined to obtain \"Air Canada\". Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible.",
"Using large corpora of chronologically ordered language, it is possible to explore diachronic phenomena, identifying previously unknown correlations between language usage and time periods, or epochs. We focused on a statistical approach to epoch delimitation and introduced the task of epoch characterization. We investigated the significant changes in the distribution of terms in the Google N-gram corpus and their relationships with emotion words. The results show that the method is reliable and the task is feasible."
]
} |
1707.08081 | 2736557444 | Search systems are often focused on providing relevant results for the "now", assuming both corpora and user needs that focus on the present. However, many corpora today reflect significant longitudinal collections ranging from 20 years of the Web to hundreds of years of digitized newspapers and books. Understanding the temporal intent of the user and retrieving the most relevant historical content has become a significant challenge. Common search features, such as query expansion, leverage the relationship between terms but cannot function well across all times when relationships vary temporally. In this work, we introduce a temporal relationship model that is extracted from longitudinal data collections. The model supports the task of identifying, given two words, when they relate to each other. We present an algorithmic framework for this task and show its application for the task of query expansion, achieving high gain. | The temporal aspects of queries and ranking gained significant attention in IR literature. Some have focused on characterizing query behavior over time. For example, different queries change in popularity over time @cite_14 and even by time of day @cite_23 . described queries to have three temporarilty patterns: atemporal, temporally unambiguous and temporally ambiguous. Others have leveraged temporal variance to look for indicators of query intent @cite_16 . Several efforts @cite_29 @cite_30 @cite_10 @cite_21 @cite_26 @cite_3 were done to not only characterize the temporal query behavior but also model it via time-series analysis. Radinsky and colleagues modeled changes in the frequency of clicked URLs, queries, and clicked query-URL pairs by using time-series analysis and show its application for improving ranking and query auto-suggestions @cite_17 @cite_11 . focused on serving the specific needs of historians, and introduced the notion of a for this purpose. | {
"cite_N": [
"@cite_30",
"@cite_14",
"@cite_26",
"@cite_11",
"@cite_29",
"@cite_21",
"@cite_3",
"@cite_23",
"@cite_16",
"@cite_10",
"@cite_17"
],
"mid": [
"2040546864",
"2127838257",
"2018022208",
"",
"2184271523",
"2165299010",
"",
"2064522604",
"2057034832",
"2057714964",
"2142076420"
],
"abstract": [
"We investigate the idea of finding semantically related search engine queries based on their temporal correlation; in other words, we infer that two queries are related if their popularities behave similarly over time. To this end, we first define a new measure of the temporal correlation of two queries based on the correlation coefficient of their frequency functions. We then conduct extensive experiments using our measure on two massive query streams from the MSN search engine, revealing that this technique can discover a wide range of semantically similar queries. Finally, we develop a method of efficiently finding the highest correlated queries for a given input query using far less space and time than the naive approach, making real-time implementation possible.",
"This project analyzed 541,920 user queries submitted to and executed in an academic Website during a four-year period (May 1997 to May 2001) using a relational data-base. The purpose of the study is three-fold: (1) to understand Web users' query behavior; (2) to identify problems encountered by these Web users; (3) to develop appropriate techniques for optimization of query analysis and mining. The linguistic analyses focus on query structures, lexicon, and word associations using statistical measures such as Zipf distribution and mutual information. A data model with finest granularity is used for data storage and iterative analyses. Patterns and trends of querying behavior are identified and compared with previous studies.",
"Seasonal events such as Halloween and Christmas repeat every year and initiate several temporal information needs. The impact of such events on users is often reflected in search logs in form of seasonal spikes in the frequency of related queries (e.g. \"halloween costumes\", \"where is santa\"). Many seasonal queries such as \"sigir conference\" mainly target fresh pages (e.g. sigir2011.org) that have less usage data such as clicks and anchor-text compared to older alternatives (e.g.sigir2009.org). Thus, it is important for search engines to correctly identify seasonal queries and make sure that their results are temporally reordered if necessary. In this poster, we focus on detecting seasonal queries using time-series analysis. We demonstrate that the seasonality of a query can be determined with high accuracy according to its historical frequency distribution.",
"",
"",
"It has become a promising direction to measure similarity of Web search queries by mining the increasing amount of click-through data logged by Web search engines, which record the interactions between users and the search engines. Most existing approaches employ the click-through data for similarity measure of queries with little consideration of the temporal factor, while the click-through data is often dynamic and contains rich temporal information. In this paper we present a new framework of time-dependent query semantic similarity model on exploiting the temporal characteristics of historical click-through data. The intuition is that more accurate semantic similarity values between queries can be obtained by taking into account the timestamps of the log data. With a set of user-defined calendar schema and calendar patterns, our time-dependent query similarity model is constructed using the marginalized kernel technique, which can exploit both explicit similarity and implicit semantics from the click-through data effectively. Experimental results on a large set of click-through data acquired from a commercial search engine show that our time-dependent query similarity model is more accurate than the existing approaches. Moreover, we observe that our time-dependent query similarity model can, to some extent, reflect real-world semantics such as real-world events that are happening over time.",
"",
"We review a query log of hundreds of millions of queries that constitute the total query traffic for an entire week of a general-purpose commercial web search service. Previously, query logs have been studied from a single, cumulative view. In contrast, our analysis shows changes in popularity and uniqueness of topically categorized queries across the hours of the day. We examine query traffic on an hourly basis by matching it against lists of queries that have been topically pre-categorized by human editors. This represents 13 of the query traffic. We show that query traffic from particular topical categories differs both from the query stream as a whole and from other categories. This analysis provides valuable insight for improving retrieval effectiveness and efficiency. It is also relevant to the development of enhanced query disambiguation, routing, and caching algorithms.",
"Web search is strongly influenced by time. The queries people issue change over time, with some queries occasionally spiking in popularity (e.g., earthquake) and others remaining relatively constant (e.g., youtube). The documents indexed by the search engine also change, with some documents always being about a particular query (e.g., the Wikipedia page on earthquakes is about the query earthquake) and others being about the query only at a particular point in time (e.g., the New York Times is only about earthquakes following a major seismic activity). The relationship between documents and queries can also change as people's intent changes (e.g., people sought different content for the query earthquake before the Haitian earthquake than they did after). In this paper, we explore how queries, their associated documents, and the query intent change over the course of 10 weeks by analyzing query log data, a daily Web crawl, and periodic human relevance judgments. We identify several interesting features by which changes to query popularity can be classified, and show that presence of these features, when accompanied by changes in result content, can be a good indicator of change in query intent.",
"We present several methods for mining knowledge from the query logs of the MSN search engine. Using the query logs, we build a time series for each query word or phrase (e.g., 'Thanksgiving' or 'Christmas gifts') where the elements of the time series are the number of times that a query is issued on a day. All of the methods we describe use sequences of this form and can be applied to time series data generally. Our primary goal is the discovery of semantically similar queries and we do so by identifying queries with similar demand patterns. Utilizing the best Fourier coefficients and the energy of the omitted components, we improve upon the state-of-the-art in time-series similarity matching. The extracted sequence features are then organized in an efficient metric tree index structure. We also demonstrate how to efficiently and accurately discover the important periods in a time-series. Finally we propose a simple but effective method for identification of bursts (long or short-term). Using the burst information extracted from a sequence, we are able to efficiently perform 'query-by-burst' on the database of time-series. We conclude the presentation with the description of a tool that uses the described methods, and serves as an interactive exploratory data discovery tool for the MSN query database.",
"The queries people issue to a search engine and the results clicked following a query change over time. For example, after the earthquake in Japan in March 2011, the query japan spiked in popularity and people issuing the query were more likely to click government-related results than they would prior to the earthquake. We explore the modeling and prediction of such temporal patterns in Web search behavior. We develop a temporal modeling framework adapted from physics and signal processing and harness it to predict temporal patterns in search behavior using smoothing, trends, periodicities, and surprises. Using current and past behavioral data, we develop a learning procedure that can be used to construct models of users' Web search activities. We also develop a novel methodology that learns to select the best prediction model from a family of predictive models for a given query or a class of queries. Experimental results indicate that the predictive models significantly outperform baseline models that weight historical evidence the same for all queries. We present two applications where new methods introduced for the temporal modeling of user behavior significantly improve upon the state of the art. Finally, we discuss opportunities for using models of temporal dynamics to enhance other areas of Web search and information retrieval."
]
} |
1707.07806 | 2737526658 | To learn a semantic parser from denotations, a learning algorithm must search over a combinatorially large space of logical forms for ones consistent with the annotated denotations. We propose a new online learning algorithm that searches faster as training progresses. The two key ideas are using macro grammars to cache the abstract patterns of useful logical forms found thus far, and holistic triggering to efficiently retrieve the most relevant patterns based on sentence similarity. On the WikiTableQuestions dataset, we first expand the search space of an existing model to improve the state-of-the-art accuracy from 38.7 to 42.7 , and then use macro grammars and holistic triggering to achieve an 11x speedup and an accuracy of 43.7 . | A traditional semantic parser maps natural language phrases into partial logical forms and composes these partial logical forms into complete logical forms. Parsers define composition based on a grammar formalism such as Combinatory Categorial Grammar (CCG) @cite_21 @cite_16 @cite_29 @cite_15 @cite_14 , Synchronous CFG @cite_23 , and CFG @cite_0 @cite_12 @cite_8 @cite_18 , while others use the syntactic structure of the utterance to guide composition @cite_27 @cite_9 . Recent neural semantic parsers allow any sequence of logical tokens to be generated @cite_5 @cite_20 @cite_24 @cite_26 @cite_6 @cite_25 . The flexibility of these composition methods allows arbitrary logical forms to be generated, but at the cost of a vastly increased search space. | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_26",
"@cite_8",
"@cite_29",
"@cite_21",
"@cite_9",
"@cite_6",
"@cite_0",
"@cite_24",
"@cite_27",
"@cite_23",
"@cite_5",
"@cite_15",
"@cite_16",
"@cite_20",
"@cite_25",
"@cite_12"
],
"mid": [
"1655078475",
"2186283982",
"2214429195",
"2252136820",
"2295690548",
"2111742432",
"2302963717",
"2950632879",
"",
"2527090203",
"2150406842",
"2121465811",
"2224454470",
"",
"2227250678",
"2444236087",
"2610403318",
"2118781169"
],
"abstract": [
"Interacting with computers is a ubiquitous activity for millions of people. Repetitive or specialized tasks often require creation of small, often one-off, programs. End-users struggle with learning and using the myriad of domain-specific languages (DSLs) to effectively accomplish these tasks. We present a general framework for constructing program synthesizers that take natural language (NL) inputs and produce expressions in a target DSL. The framework takes as input a DSL definition and training data consisting of NL DSL pairs. From these it constructs a synthesizer by learning optimal weights and classifiers (using NLP features) that rank the outputs of a keyword-programming based translation. We applied our framework to three domains: repetitive text editing, an intelligent tutoring system, and flight information queries. On 1200+ English descriptions, the respective synthesizers rank the desired program as the top-1 and top-3 for 80 and 90 descriptions respectively.",
"This paper introduces Logical Semantics with Perception (LSP), a model for grounded language acquisition that learns to map natural language statements to their referents in a physical environment. For example, given an image, LSP can map the statement “blue mug on the table” to the set of image segments showing blue mugs on tables. LSP learns physical representations for both categorical (“blue,” “mug”) and relational (“on”) language, and also learns to compose these representations to produce the referents of entire statements. We further introduce a weakly supervised training procedure that estimates LSP’s parameters using annotated referents for entire statements, without annotated referents for individual words or the parse structure of the statement. We perform experiments on two applications: scene understanding and geographical question answering. We find that LSP outperforms existing, less expressive models that cannot represent relational language. We further find that weakly supervised training is competitive with fully supervised training while requiring significantly less annotation effort.",
"Deep neural networks have achieved impressive supervised classification performance in many tasks including image recognition, speech recognition, and sequence to sequence learning. However, this success has not been translated to applications like question answering that may involve complex arithmetic and logic reasoning. A major limitation of these models is in their inability to learn even simple arithmetic and logic operations. For example, it has been shown that neural networks fail to learn to add two binary numbers reliably. In this work, we propose Neural Programmer, an end-to-end differentiable neural network augmented with a small set of basic arithmetic and logic operations. Neural Programmer can call these augmented operations over several steps, thereby inducing compositional programs that are more complex than the built-in operations. The model learns from a weak supervision signal which is the result of execution of the correct program, hence it does not require expensive annotation of the correct program itself. The decisions of what operations to call, and what data segments to apply to are inferred by Neural Programmer. Such decisions, during training, are done in a differentiable fashion so that the entire network can be trained jointly by gradient descent. We find that training the model is difficult, but it can be greatly improved by adding random noise to the gradient. On a fairly complex synthetic table-comprehension dataset, traditional recurrent networks and attentional models perform poorly while Neural Programmer typically obtains nearly perfect accuracy.",
"In this paper, we train a semantic parser that scales up to Freebase. Instead of relying on annotated logical forms, which is especially expensive to obtain at large scale, we learn from question-answer pairs. The main challenge in this setting is narrowing down the huge number of possible logical predicates for a given question. We tackle this problem in two ways: First, we build a coarse mapping from phrases to predicates using a knowledge base and a large text corpus. Second, we use a bridging operation to generate additional predicates based on neighboring predicates. On the dataset of Cai and Yates (2013), despite not having annotated logical forms, our system outperforms their state-of-the-art parser. Additionally, we collected a more realistic and challenging dataset of question-answer pairs and improves over a natural baseline.",
"Semantic parsers conventionally construct logical forms bottom-up in a fixed order, resulting in the generation of many extraneous partial logical forms. In this paper, we combine ideas from imitation learning and agenda-based parsing to train a semantic parser that searches partial logical forms in a more strategic order. Empirically, our parser reduces the number of constructed partial logical forms by an order of magnitude, and obtains a 6x-9x speedup over fixed-order parsing, while maintaining comparable accuracy.",
"We consider the problem of learning to parse sentences to lambda-calculus representations of their underlying semantics and present an algorithm that learns a weighted combinatory categorial grammar (CCG). A key idea is to introduce non-standard CCG combinators that relax certain parts of the grammar—for example allowing flexible word order, or insertion of lexical items— with learned costs. We also present a new, online algorithm for inducing a weighted CCG. Results for the approach on ATIS data show 86 F-measure in recovering fully correct semantic analyses and 95.9 F-measure by a partial-match criterion, a more than 5 improvement over the 90.3 partial-match figure reported by He and Young (2006).",
"The strongly typed syntax of grammar formalisms such as CCG, TAG, LFG and HPSG offers a synchronous framework for deriving syntactic structures and semantic logical forms. In contrast---partly due to the lack of a strong type system---dependency structures are easy to annotate and have become a widely used form of syntactic analysis for many languages. However, the lack of a type system makes a formal mechanism for deriving logical forms from dependency structures challenging. We address this by introducing a robust system based on the lambda calculus for deriving neo-Davidsonian logical forms from dependency trees. These logical forms are then used for semantic parsing of natural language to Freebase. Experiments on the Free917 and WebQuestions datasets show that our representation is superior to the original dependency trees and that it outperforms a CCG-based representation on this task. Compared to prior work, we obtain the strongest result to date on Free917 and competitive results on WebQuestions.",
"Harnessing the statistical power of neural networks to perform language understanding and symbolic reasoning is difficult, when it requires executing efficient discrete operations against a large knowledge-base. In this work, we introduce a Neural Symbolic Machine, which contains (a) a neural \"programmer\", i.e., a sequence-to-sequence model that maps language utterances to programs and utilizes a key-variable memory to handle compositionality (b) a symbolic \"computer\", i.e., a Lisp interpreter that performs program execution, and helps find good programs by pruning the search space. We apply REINFORCE to directly optimize the task reward of this structured prediction problem. To train with weak supervision and improve the stability of REINFORCE, we augment it with an iterative maximum-likelihood training process. NSM outperforms the state-of-the-art on the WebQuestionsSP dataset when trained from question-answer pairs only, without requiring any feature engineering or domain-specific knowledge.",
"",
"We present a novel semi-supervised approach for sequence transduction and apply it to semantic parsing. The unsupervised component is based on a generative model in which latent sentences generate the unpaired logical forms. We apply this method to a number of semantic parsing tasks focusing on domains with limited access to labelled training data and extend those datasets with synthetically generated logical forms.",
"We present the first unsupervised approach to the problem of learning a semantic parser, using Markov logic. Our USP system transforms dependency trees into quasi-logical forms, recursively induces lambda forms from these, and clusters them to abstract away syntactic variations of the same meaning. The MAP semantic parse of a sentence is obtained by recursively assigning its parts to lambda-form clusters and composing them. We evaluate our approach by using it to extract a knowledge base from biomedical abstracts and answer questions. USP substantially outperforms TextRunner, DIRT and an informed baseline on both precision and recall on this task.",
"This paper presents the first empirical results to our knowledge on learning synchronous grammars that generate logical forms. Using statistical machine translation techniques, a semantic parser based on a synchronous context-free grammar augmented with �operators is learned given a set of training sentences and their correct logical forms. The resulting parser is shown to be the bestperforming system so far in a database query domain.",
"Semantic parsing aims at mapping natural language to machine interpretable meaning representations. Traditional approaches rely on high-quality lexicons, manually-built templates, and linguistic features which are either domain- or representation-specific. In this paper we present a general method based on an attention-enhanced encoder-decoder model. We encode input utterances into vector representations, and generate their logical forms by conditioning the output sequences or trees on the encoding vectors. Experimental results on four datasets show that our approach performs competitively without using hand-engineered features and is easy to adapt across domains and meaning representations.",
"",
"We consider the problem of learning factored probabilistic CCG grammars for semantic parsing from data containing sentences paired with logical-form meaning representations. Traditional CCG lexicons list lexical items that pair words and phrases with syntactic and semantic content. Such lexicons can be inefficient when words appear repeatedly with closely related lexical content. In this paper, we introduce factored lexicons, which include both lexemes to model word meaning and templates to model systematic variation in word usage. We also present an algorithm for learning factored CCG lexicons, along with a probabilistic parse-selection model. Evaluations on benchmark datasets demonstrate that the approach learns highly accurate parsers, whose generalization performance benefits greatly from the lexical factoring.",
"Modeling crisp logical regularities is crucial in semantic parsing, making it difficult for neural models with no task-specific prior knowledge to achieve good results. In this paper, we introduce data recombination, a novel framework for injecting such prior knowledge into a model. From the training data, we induce a high-precision synchronous context-free grammar, which captures important conditional independence properties commonly found in semantic parsing. We then train a sequence-to-sequence recurrent network (RNN) model with a novel attention-based copying mechanism on datapoints sampled from this grammar, thereby teaching the model about these structural properties. Data recombination improves the accuracy of our RNN model on three semantic parsing datasets, leading to new state-of-the-art performance on the standard GeoQuery dataset for models with comparable supervision.",
"Our goal is to learn a semantic parser that maps natural language utterances into executable programs when only indirect supervision is available: examples are labeled with the correct execution result, but not the program itself. Consequently, we must search the space of programs for those that output the correct result, while not being misled by spurious programs: incorrect programs that coincidentally output the correct result. We connect two common learning paradigms, reinforcement learning (RL) and maximum marginal likelihood (MML), and then present a new learning algorithm that combines the strengths of both. The new algorithm guards against spurious programs by combining the systematic search traditionally employed in MML with the randomized exploration of RL, and by updating parameters such that probability is spread more evenly across consistent programs. We apply our learning algorithm to a new neural semantic parser and show significant gains over existing state-of-the-art results on a recent context-dependent semantic parsing task.",
"The ability to understand natural-language instructions is critical to building intelligent agents that interact with humans. We present a system that learns to transform natural-language navigation instructions into executable formal plans. Given no prior linguistic knowledge, the system learns by simply observing how humans follow navigation instructions. The system is evaluated in three complex virtual indoor environments with numerous objects and landmarks. A previously collected realistic corpus of complex English navigation instructions for these environments is used for training and testing data. By using a learned lexicon to refine inferred plans and a supervised learner to induce a semantic parser, the system is able to automatically learn to correctly interpret a reasonable fraction of the complex instructions in this corpus."
]
} |
1707.07699 | 2952152857 | In this paper, we discuss the feasibility of monitoring partially synchronous distributed systems to detect latent bugs, i.e., errors caused by concurrency and race conditions among concurrent processes. We present a monitoring framework where we model both system constraints and latent bugs as Satisfiability Modulo Theories (SMT) formulas, and we detect the presence of latent bugs using an SMT solver. We demonstrate the feasibility of our framework using both synthetic applications where latent bugs occur at any time with random probability and an application involving exclusive access to a shared resource with a subtle timing bug. We illustrate how the time required for verification is affected by parameters such as communication frequency, latency, and clock skew. Our results show that our framework can be used for real-life applications, and because our framework uses SMT solvers, the range of appropriate applications will increase as these solvers become more efficient over time. | The fundamental challenge in distributed predicate detection lies in causality induced by inherent non-determinism @cite_4 . Most existing distributed system monitoring works have focused on asynchronous systems that use vector clocks (VCs) @cite_2 @cite_19 and @cite_9 @cite_7 @cite_10 @cite_11 @cite_8 make minimal assumptions about the underlying system. Unfortunately, asynchronous monitors have several sources of inefficiency that limit their scalability and impede their adoption in real systems. First, general predicate detection with asynchronous monitors is NP-complete @cite_11 , although polynomial time algorithms exists for special cases of predicate detection such as linear predicates and conjunctive predicates @cite_11 @cite_9 . Second, asynchronous monitors require @math space VC timestamps to track all causalities @cite_8 in the system where @math is the number of processes. Thus, each message has linear size in the system which limits scalability of asynchronous monitors. | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_9",
"@cite_19",
"@cite_2",
"@cite_10",
"@cite_11"
],
"mid": [
"1981173460",
"",
"2071933194",
"2160800498",
"1652742168",
"2116746874",
"2016720263",
"2052995544"
],
"abstract": [
"The paper shows that characterizing the causal relationship between significant events is an important but non-trivial aspect for understanding the behavior of distributed programs. An introduction to the notion of causality and its relation to logical time is given; some fundamental results concerning the characterization of causality are presented. Recent work on the detection of causal relationships in distributed computations is surveyed. The issue of observing distributed computations in a causally consistent way and the basic problems of detecting global predicates are discussed. To illustrate the major difficulties, some typical monitoring and debugging approaches are assessed, and it is demonstrated how their feasibility is severely limited by the fundamental problem to master the complexity of causal relationships.",
"",
"",
"This paper discusses detection of global predicates in a distributed program. Earlier algorithms for detection of global predicates proposed by Chandy and Lamport (1985) work only for stable predicates. A predicate is stable if it does not turn false once it becomes true. Our algorithms detect even unstable predicates, without excessive overhead. In the past, such predicates have been regarded as too difficult to detect. The predicates are specified by using a logic described formally in this paper. We discuss detection of weak conjunctive predicates that are formed by conjunction of predicates local to processes in the system. Our detection methods will detect whether such a predicate is true for any interleaving of events in the system, regardless of whether the predicate is stable. Also, any predicate that can be reduced to a set of weak conjunctive predicates is detectable. This class of predicates captures many global predicates that are of interest to a programmer. The message complexity of our algorithm is bounded by the number of messages used by the program. The main applications of our results are in debugging and testing of distributed programs. Our algorithms have been incorporated in a distributed debugger that runs on a network of Sun workstations in UNIX. >",
"A distributed system can be characterized by the fact that the global state is distributed and that a common time base does not exist. However, the notion of time is an important concept in every day life of our decentralized world\" and helps to solve problems like getting a consistent population census or determining the potential causality between events. We argue that a linearly ordered structure of time is not (always) adequate for distributed systems and propose a generalized non-standard model of time which consists of vectors of clocks. These clock-vectors are partially ordered and form a lattice. By using timestamps and a simple clock update mechanism the structure of causality is represented in an isomorphic way. The new model of time has a close analogy to Minkowski's relativistic spacetime and leads among others to an interesting characterization of the global state problem. Finally, we present a new algorithm to compute a consistent global snapshot of a distributed system where messages may be received out of order.",
"Timestamping is a common method of totally ordering events in concurrent programs. However, for applications requiring access to the global state, a total ordering is inappro priate. This paper presents algorithms for timestamping events in both synchronous and asynchronous n1essage-passing programs that allow for access to the partial ordering in herent in a parallel system. The algorithms do not change the con1munications graph or require a central timestamp issuing authority.",
"In an asynchronous distributed system, a number of processes communicate with each other via message passing that has a finite but arbitrary long delay. There is no global clock in that system. Predicates, denoting the states of processes and their relations, are often used to specify the information of interest in such a system. Due to the lack of a global clock, the temporal relations between the states at different processes cannot be uniquely determined, but have multiple possible circumstances. Existing works of predicate detection are based on the definitely modality or the possibly modality, denoting that a predicate holds in all of the possible circumstances or in one of them, respectively. No information is provided about the probability that a predicate will hold, which hinders the taking of countermeasures for different situations. Moreover, the detection is based on single occurrence of a predicate, so the results are heavily affected by environmental noise and detection errors. In this paper, we propose a new approach to predicate detection to address these two issues. We generalize the definitely and possibly modalities to an occurrence probability to provide more detailed information, and further investigate how to detect multiple occurrences of a predicate. We propose a unified algorithm framework for detecting various types of predicates and demonstrate the use of it for three typical types of predicates, including simple predicates, simple sequences, and interval-constrained sequences. Theoretical proofs and simulation results show that our approach is effective and outperforms existing approaches.",
"We show that the problem of predicate detection in distributed systems is NP-complete. In the past, efficient algorithms have been developed for special classes of predicates such as stable predicates, observer independent predicates, and conjunctive predicates. We introduce a class of predicates, semi-linear predicates, which properly contains all of the above classes. We first discuss stable, observer independent and semi-linear classes of predicates and their relationships with each other. We also study closure properties of these classes with respect to conjunction and disjunction. Finally, we discuss algorithms for detection of predicates in these classes. We provide a nondeterministic detection algorithm for each class of predicate. We show that each class can be equivalently characterized by the degree of non-determinism present in the algorithm. Stable predicates are defined as those that can be detected by an algorithm with the most non-determinism. All other classes can be derived by appropriately constraining the non-determinism in this algorithm."
]
} |
1707.08063 | 2737988395 | We study the problem of estimating the relative depth order of point pairs in a monocular image. Recent advances mainly focus on using deep convolutional neural networks (DCNNs) to learn and infer the ordinal information from multiple contextual information of the points pair such as global scene context, local contextual information, and the locations. However, it remains unclear how much each context contributes to the task. To address this, we first examine the contribution of each context cue [1], [2] to the performance in the context of depth order estimation. We find out the local context surrounding the points pair contributes the most and the global scene context helps little. Based on the findings, we propose a simple method, using a multi-scale densely-connected network to tackle the task. Instead of learning the global structure, we dedicate to explore the local structure by learning to regress from regions of multiple sizes around the point pairs. Moreover, we use the recent densely connected network [3] to encourage substantial feature reuse as well as deepen our network to boost the performance. We show in experiments that the results of our approach is on par with or better than the state-of-the-art methods with the benefit of using only a small number of training data. | Recently, due to the wide adoption of affordable depth sensors, datasets such as the NYU Depth dataset @cite_5 and KITTI dataset @cite_33 @cite_10 become available, thus leading to the trend of solving the problem as a supervised learning task using CNN. Pioneering work of @cite_7 proposed an end-to-end system to estimate the depth order of points pairs. Compared with another ordinal estimation work which also utilized CNN @cite_15 , add one more bounding box which contains the important visual cues for the task and a different location expression in their network structure. We consider such the design very practical, which makes the visual cues be automatically learned and inferred by CNN. Since the NYU and KITTI datasets have no direct annotations for the task, @cite_6 proposed a dataset named Depth in the Wild (DIW) which annotated the ordinal depth information between point pairs. Their deep model is an deep ranking model, and the output is a relative depth map. | {
"cite_N": [
"@cite_33",
"@cite_7",
"@cite_6",
"@cite_5",
"@cite_15",
"@cite_10"
],
"mid": [
"2115579991",
"2221366145",
"2339763956",
"125693051",
"",
""
],
"abstract": [
"We present a novel dataset captured from a VW station wagon for use in mobile robotics and autonomous driving research. In total, we recorded 6 hours of traffic scenarios at 10-100 Hz using a variety of sensor modalities such as high-resolution color and grayscale stereo cameras, a Velodyne 3D laser scanner and a high-precision GPS IMU inertial navigation system. The scenarios are diverse, capturing real-world traffic situations, and range from freeways over rural areas to inner-city scenes with many static and dynamic objects. Our data is calibrated, synchronized and timestamped, and we provide the rectified and raw image sequences. Our dataset also contains object labels in the form of 3D tracklets, and we provide online benchmarks for stereo, optical flow, object detection and other tasks. This paper describes our recording platform, the data format and the utilities that we provide.",
"We propose a framework that infers mid-level visual properties of an image by learning about ordinal relationships. Instead of estimating metric quantities directly, the system proposes pairwise relationship estimates for points in the input image. These sparse probabilistic ordinal measurements are globalized to create a dense output map of continuous metric measurements. Estimating order relationships between pairs of points has several advantages over metric estimation: it solves a simpler problem than metric regression, humans are better at relative judgements, so data collection is easier, ordinal relationships are invariant to monotonic transformations of the data, thereby increasing the robustness of the system and providing qualitatively different information. We demonstrate that this frame-work works well on two important mid-level vision tasks: intrinsic image decomposition and depth from an RGB image. We train two systems with the same architecture on data from these two modalities. We provide an analysis of the resulting models, showing that they learn a number of simple rules to make ordinal decisions. We apply our algorithm to depth estimation, with good results, and intrinsic image decomposition, with state-of-the-art results.",
"This paper studies single-image depth perception in the wild, i.e., recovering depth from a single image taken in unconstrained settings. We introduce a new dataset \"Depth in the Wild\" consisting of images in the wild annotated with relative depth between pairs of random points. We also propose a new algorithm that learns to estimate metric depth using annotations of relative depth. Compared to the state of the art, our algorithm is simpler and performs better. Experiments show that our algorithm, combined with existing RGB-D data and our new relative depth annotations, significantly improves single-image depth perception in the wild.",
"We present an approach to interpret the major surfaces, objects, and support relations of an indoor scene from an RGBD image. Most existing work ignores physical interactions or is applied only to tidy rooms and hallways. Our goal is to parse typical, often messy, indoor scenes into floor, walls, supporting surfaces, and object regions, and to recover support relationships. One of our main interests is to better understand how 3D cues can best inform a structured 3D interpretation. We also contribute a novel integer programming formulation to infer physical support relations. We offer a new dataset of 1449 RGBD images, capturing 464 diverse indoor scenes, with detailed annotations. Our experiments demonstrate our ability to infer support relations in complex scenes and verify that our 3D scene cues and inferred support lead to better object segmentation.",
"",
""
]
} |
1707.08063 | 2737988395 | We study the problem of estimating the relative depth order of point pairs in a monocular image. Recent advances mainly focus on using deep convolutional neural networks (DCNNs) to learn and infer the ordinal information from multiple contextual information of the points pair such as global scene context, local contextual information, and the locations. However, it remains unclear how much each context contributes to the task. To address this, we first examine the contribution of each context cue [1], [2] to the performance in the context of depth order estimation. We find out the local context surrounding the points pair contributes the most and the global scene context helps little. Based on the findings, we propose a simple method, using a multi-scale densely-connected network to tackle the task. Instead of learning the global structure, we dedicate to explore the local structure by learning to regress from regions of multiple sizes around the point pairs. Moreover, we use the recent densely connected network [3] to encourage substantial feature reuse as well as deepen our network to boost the performance. We show in experiments that the results of our approach is on par with or better than the state-of-the-art methods with the benefit of using only a small number of training data. | The densely connected network is a new deep structure proposed by @cite_38 . It takes advantage of the above useful designs such as small filters ( @math , @math ), block by block structure (yet apply average pooling) and skip connection as well. In particular, its primary characteristic is the dense connectivity, which leads to the heavy feature reuse, information flow propagation, and good regularization. Considering these advantages, we apply the DenseNet in our proposed model. To verify our choice, we compare its effectiveness against ResNet and the baseline model, respectively. We observe the improvement over the proposed baseline model. The details are shown in Section . | {
"cite_N": [
"@cite_38"
],
"mid": [
"2511730936"
],
"abstract": [
"Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections - one between each layer and its subsequent layer - our network has L(L+1) 2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less computation to achieve high performance. Code and pre-trained models are available at this https URL ."
]
} |
1707.07998 | 2951590222 | Top-down visual attention mechanisms have been used extensively in image captioning and visual question answering (VQA) to enable deeper image understanding through fine-grained analysis and even multiple steps of reasoning. In this work, we propose a combined bottom-up and top-down attention mechanism that enables attention to be calculated at the level of objects and other salient image regions. This is the natural basis for attention to be considered. Within our approach, the bottom-up mechanism (based on Faster R-CNN) proposes image regions, each with an associated feature vector, while the top-down mechanism determines feature weightings. Applying this approach to image captioning, our results on the MSCOCO test server establish a new state-of-the-art for the task, achieving CIDEr SPICE BLEU-4 scores of 117.9, 21.5 and 36.9, respectively. Demonstrating the broad applicability of the method, applying the same approach to VQA we obtain first place in the 2017 VQA Challenge. | A large number of attention-based deep neural networks have been proposed for image captioning and VQA. Typically, these models can be characterized as top-down approaches, with context provided by a representation of a partially-completed caption in the case of image captioning @cite_6 @cite_40 @cite_20 @cite_48 , or a representation of the question in the case of VQA @cite_41 @cite_34 @cite_3 @cite_2 @cite_46 . In each case attention is applied to the output of one or more layers of a CNN, by predicting a weighting for each spatial location in the CNN output. However, determining the optimal number of image regions invariably requires an unwinnable trade-off between coarse and fine levels of detail. Furthermore, the arbitrary positioning of the regions with respect to image content may make it more difficult to detect objects that are poorly aligned to regions and to bind visual concepts associated with the same object. | {
"cite_N": [
"@cite_41",
"@cite_48",
"@cite_6",
"@cite_3",
"@cite_40",
"@cite_2",
"@cite_46",
"@cite_34",
"@cite_20"
],
"mid": [
"2963383024",
"2950178297",
"2963084599",
"2255577267",
"2952469094",
"2171810632",
"2136462581",
"2963668159",
"2544271936"
],
"abstract": [
"",
"Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr8k, Flickr30k and MS COCO.",
"Recently it has been shown that policy-gradient methods for reinforcement learning can be utilized to train deep end-to-end systems directly on non-differentiable metrics for the task at hand. In this paper we consider the problem of optimizing image captioning systems using reinforcement learning, and show that by carefully optimizing our systems using the test metrics of the MSCOCO task, significant gains in performance can be realized. Our systems are built using a new optimization approach that we call self-critical sequence training (SCST). SCST is a form of the popular REINFORCE algorithm that, rather than estimating a baseline to normalize the rewards and reduce variance, utilizes the output of its own test-time inference algorithm to normalize the rewards it experiences. Using this approach, estimating the reward signal (as actor-critic methods must do) and estimating normalization (as REINFORCE algorithms typically do) is avoided, while at the same time harmonizing the model with respect to its test-time inference procedure. Empirically we find that directly optimizing the CIDEr metric with SCST and greedy decoding at test-time is highly effective. Our results on the MSCOCO evaluation sever establish a new state-of-the-art on the task, improving the best result in terms of CIDEr from 104.9 to 114.7.",
"We address the problem of Visual Question Answering (VQA), which requires joint image and language understanding to answer a question about a given photograph. Recent approaches have applied deep image captioning methods based on convolutional-recurrent networks to this problem, but have failed to model spatial inference. To remedy this, we propose a model we call the Spatial Memory Network and apply it to the VQA task. Memory networks are recurrent neural networks with an explicit attention mechanism that selects certain parts of the information stored in memory. Our Spatial Memory Network stores neuron activations from different spatial regions of the image in its memory, and uses the question to choose relevant regions for computing the answer, a process of which constitutes a single \"hop\" in the network. We propose a novel spatial attention architecture that aligns words with image patches in the first hop, and obtain improved results by adding a second attention hop which considers the whole question to choose visual evidence based on the results of the first hop. To better understand the inference process learned by the network, we design synthetic questions that specifically require spatial inference and visualize the attention weights. We evaluate our model on two published visual question answering datasets, DAQUAR [1] and VQA [2], and obtain improved results compared to a strong deep baseline model (iBOWIMG) which concatenates image and question features to predict the answer [3].",
"Attention-based neural encoder-decoder frameworks have been widely adopted for image captioning. Most methods force visual attention to be active for every generated word. However, the decoder likely requires little to no visual information from the image to predict non-visual words such as \"the\" and \"of\". Other words that may seem visual can often be predicted reliably just from the language model e.g., \"sign\" after \"behind a red stop\" or \"phone\" following \"talking on a cell\". In this paper, we propose a novel adaptive attention model with a visual sentinel. At each time step, our model decides whether to attend to the image (and if so, to which regions) or to the visual sentinel. The model decides whether to attend to the image and where, in order to extract meaningful information for sequential word generation. We test our method on the COCO image captioning 2015 challenge dataset and Flickr30K. Our approach sets the new state-of-the-art by a significant margin.",
"This paper presents stacked attention networks (SANs) that learn to answer natural language questions from images. SANs use semantic representation of a question as query to search for the regions in an image that are related to the answer. We argue that image question answering (QA) often requires multiple steps of reasoning. Thus, we develop a multiple-layer SAN in which we query an image multiple times to infer the answer progressively. Experiments conducted on four image QA data sets demonstrate that the proposed SANs significantly outperform previous state-of-the-art approaches. The visualization of the attention layers illustrates the progress that the SAN locates the relevant visual clues that lead to the answer of the question layer-by-layer.",
"We have seen great progress in basic perceptual tasks such as object recognition and detection. However, AI models still fail to match humans in high-level vision tasks due to the lack of capacities for deeper reasoning. Recently the new task of visual question answering (QA) has been proposed to evaluate a model’s capacity for deep image understanding. Previous works have established a loose, global association between QA sentences and images. However, many questions and answers, in practice, relate to local regions in the images. We establish a semantic link between textual descriptions and image regions by object-level grounding. It enables a new type of QA with visual answers, in addition to textual answers used in previous work. We study the visual QA tasks in a grounded setting with a large collection of 7W multiple-choice QA pairs. Furthermore, we evaluate human performance and several baseline models on the QA tasks. Finally, we propose a novel LSTM model with spatial attention to tackle the 7W QA tasks.",
"A number of recent works have proposed attention models for Visual Question Answering (VQA) that generate spatial maps highlighting image regions relevant to answering the question. In this paper, we argue that in addition to modeling \"where to look\" or visual attention, it is equally important to model \"what words to listen to\" or question attention. We present a novel co-attention model for VQA that jointly reasons about image and question attention. In addition, our model reasons about the question (and consequently the image via the co-attention mechanism) in a hierarchical fashion via a novel 1-dimensional convolution neural networks (CNN). Our model improves the state-of-the-art on the VQA dataset from 60.3 to 60.5 , and from 61.6 to 63.3 on the COCO-QA dataset. By using ResNet, the performance is further improved to 62.1 for VQA and 65.4 for COCO-QA.",
"We propose a novel extension of the encoder-decoder framework, called a review network. The review network is generic and can enhance any existing encoder- decoder model: in this paper, we consider RNN decoders with both CNN and RNN encoders. The review network performs a number of review steps with attention mechanism on the encoder hidden states, and outputs a thought vector after each review step; the thought vectors are used as the input of the attention mechanism in the decoder. We show that conventional encoder-decoders are a special case of our framework. Empirically, we show that our framework improves over state-of- the-art encoder-decoder systems on the tasks of image captioning and source code captioning."
]
} |
1707.07998 | 2951590222 | Top-down visual attention mechanisms have been used extensively in image captioning and visual question answering (VQA) to enable deeper image understanding through fine-grained analysis and even multiple steps of reasoning. In this work, we propose a combined bottom-up and top-down attention mechanism that enables attention to be calculated at the level of objects and other salient image regions. This is the natural basis for attention to be considered. Within our approach, the bottom-up mechanism (based on Faster R-CNN) proposes image regions, each with an associated feature vector, while the top-down mechanism determines feature weightings. Applying this approach to image captioning, our results on the MSCOCO test server establish a new state-of-the-art for the task, achieving CIDEr SPICE BLEU-4 scores of 117.9, 21.5 and 36.9, respectively. Demonstrating the broad applicability of the method, applying the same approach to VQA we obtain first place in the 2017 VQA Challenge. | Comparatively few previous works have considered applying attention to salient image regions. We are aware of two papers. @cite_26 use selective search @cite_18 to identify salient image regions, which are filtered with a classifier then resized and CNN-encoded as input to an image captioning model with attention. The Areas of Attention captioning model @cite_30 uses either edge boxes @cite_7 or spatial transformer networks @cite_32 to generate image features, which are processed using an attention model based on three bi-linear pairwise interactions @cite_30 . In this work, rather than using hand-crafted or differentiable region proposals @cite_18 @cite_7 @cite_32 , we leverage Faster R-CNN @cite_42 , establishing a closer link between vision and language tasks and recent progress in object detection. With this approach we are able to pre-train our region proposals on object detection datasets. Conceptually, the advantages should be similar to pre-training visual representations on ImageNet @cite_8 and leveraging significantly larger cross-domain knowledge. We additionally apply our method to VQA, establishing the broad applicability of our approach. | {
"cite_N": [
"@cite_30",
"@cite_18",
"@cite_26",
"@cite_7",
"@cite_8",
"@cite_42",
"@cite_32"
],
"mid": [
"2949906242",
"2088049833",
"1785460851",
"2949769367",
"2952020226",
"2953106684",
"2951005624"
],
"abstract": [
"We propose \"Areas of Attention\", a novel attention-based model for automatic image captioning. Our approach models the dependencies between image regions, caption words, and the state of an RNN language model, using three pairwise interactions. In contrast to previous attention-based approaches that associate image regions only to the RNN state, our method allows a direct association between caption words and image regions. During training these associations are inferred from image-level captions, akin to weakly-supervised object detector training. These associations help to improve captioning by localizing the corresponding regions during testing. We also propose and compare different ways of generating attention areas: CNN activation grids, object proposals, and spatial transformers nets applied in a convolutional fashion. Spatial transformers give the best results. They allow for image specific attention areas, and can be trained jointly with the rest of the network. Our attention mechanism and spatial transformer attention areas together yield state-of-the-art results on the MSCOCO dataset.o meaningful latent semantic structure in the generated captions.",
"This paper addresses the problem of generating possible object locations for use in object recognition. We introduce selective search which combines the strength of both an exhaustive search and segmentation. Like segmentation, we use the image structure to guide our sampling process. Like exhaustive search, we aim to capture all possible object locations. Instead of a single technique to generate possible object locations, we diversify our search and use a variety of complementary image partitionings to deal with as many image conditions as possible. Our selective search results in a small set of data-driven, class-independent, high quality locations, yielding 99 recall and a Mean Average Best Overlap of 0.879 at 10,097 locations. The reduced number of locations compared to an exhaustive search enables the use of stronger machine learning techniques and stronger appearance models for object recognition. In this paper we show that our selective search enables the use of the powerful Bag-of-Words model for recognition. The selective search software is made publicly available (Software: http: disi.unitn.it uijlings SelectiveSearch.html ).",
"Recent progress on automatic generation of image captions has shown that it is possible to describe the most salient information conveyed by images with accurate and meaningful sentences. In this paper, we propose an image caption system that exploits the parallel structures between images and sentences. In our model, the process of generating the next word, given the previously generated ones, is aligned with the visual perception experience where the attention shifting among the visual regions imposes a thread of visual ordering. This alignment characterizes the flow of \"abstract meaning\", encoding what is semantically shared by both the visual scene and the text description. Our system also makes another novel modeling contribution by introducing scene-specific contexts that capture higher-level semantic information encoded in an image. The contexts adapt language models for word generation to specific scene types. We benchmark our system and contrast to published results on several popular datasets. We show that using either region-based attention or scene-specific contexts improves systems without those components. Furthermore, combining these two modeling ingredients attains the state-of-the-art performance.",
"This paper presents a novel approach for automatically generating image descriptions: visual detectors, language models, and multimodal similarity models learnt directly from a dataset of image captions. We use multiple instance learning to train visual detectors for words that commonly occur in captions, including many different parts of speech such as nouns, verbs, and adjectives. The word detector outputs serve as conditional inputs to a maximum-entropy language model. The language model learns from a set of over 400,000 image descriptions to capture the statistics of word usage. We capture global semantics by re-ranking caption candidates using sentence-level features and a deep multimodal similarity model. Our system is state-of-the-art on the official Microsoft COCO benchmark, producing a BLEU-4 score of 29.1 . When human judges compare the system captions to ones written by other people on our held-out test set, the system captions have equal or better quality 34 of the time.",
"The ImageNet Large Scale Visual Recognition Challenge is a benchmark in object category classification and detection on hundreds of object categories and millions of images. The challenge has been run annually from 2010 to present, attracting participation from more than fifty institutions. This paper describes the creation of this benchmark dataset and the advances in object recognition that have been possible as a result. We discuss the challenges of collecting large-scale ground truth annotation, highlight key breakthroughs in categorical object recognition, provide a detailed analysis of the current state of the field of large-scale image classification and object detection, and compare the state-of-the-art computer vision accuracy with human accuracy. We conclude with lessons learned in the five years of the challenge, and propose future directions and improvements.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.",
"Convolutional Neural Networks define an exceptionally powerful class of models, but are still limited by the lack of ability to be spatially invariant to the input data in a computationally and parameter efficient manner. In this work we introduce a new learnable module, the Spatial Transformer, which explicitly allows the spatial manipulation of data within the network. This differentiable module can be inserted into existing convolutional architectures, giving neural networks the ability to actively spatially transform feature maps, conditional on the feature map itself, without any extra training supervision or modification to the optimisation process. We show that the use of spatial transformers results in models which learn invariance to translation, scale, rotation and more generic warping, resulting in state-of-the-art performance on several benchmarks, and for a number of classes of transformations."
]
} |
1707.07791 | 2736538122 | Learning the distance metric between pairs of examples is of great importance for visual recognition, especially for person re-identification (Re-Id). Recently, the contrastive and triplet loss are proposed to enhance the discriminative power of the deeply learned features, and have achieved remarkable success. As can be seen, either the contrastive or triplet loss is just one special case of the Euclidean distance relationships among these training samples. Therefore, we propose a structured graph Laplacian embedding algorithm, which can formulate all these structured distance relationships into the graph Laplacian form. The proposed method can take full advantages of the structured distance relationships among these training samples, with the constructed complete graph. Besides, this formulation makes our method easy-to-implement and super-effective. When embedding the proposed algorithm with the softmax loss for the CNN training, our method can obtain much more robust and discriminative deep features with inter-personal dispersion and intra-personal compactness, which is essential to person Re-Id. We illustrate the effectiveness of our proposed method on top of three popular networks, namely AlexNet, DGDNet and ResNet50, on recent four widely used Re-Id benchmark datasets. Our proposed method achieves state-of-the-art performances. | A large number of metric learning and ranking algorithms have been applied to the person Re-Id problem @cite_0 @cite_22 . The basic idea behind metric learning is to find a mapping function from feature space to the distance space with certain merits, such as the feature representations from the same person being closer than those from different ones @cite_29 . These metric learning methods mainly include the Mahalanobias metric learning @cite_35 , local fisher analysis @cite_40 , large margin nearest neighbour @cite_40 , locally adaptive decision functions @cite_23 , attribute based consistent matching @cite_48 , the polynomial kernel method @cite_60 and the XQDA algorithm @cite_25 . | {
"cite_N": [
"@cite_35",
"@cite_22",
"@cite_60",
"@cite_48",
"@cite_29",
"@cite_0",
"@cite_40",
"@cite_23",
"@cite_25"
],
"mid": [
"2068042582",
"",
"2475284720",
"",
"",
"166429404",
"2106053110",
"",
"1949591461"
],
"abstract": [
"In this paper, we raise important issues on scalability and the required degree of supervision of existing Mahalanobis metric learning methods. Often rather tedious optimization procedures are applied that become computationally intractable on a large scale. Further, if one considers the constantly growing amount of data it is often infeasible to specify fully supervised labels for all data points. Instead, it is easier to specify labels in form of equivalence constraints. We introduce a simple though effective strategy to learn a distance metric from equivalence constraints, based on a statistical inference perspective. In contrast to existing methods we do not rely on complex optimization problems requiring computationally expensive iterations. Hence, our method is orders of magnitudes faster than comparable methods. Results on a variety of challenging benchmarks with rather diverse nature demonstrate the power of our method. These include faces in unconstrained environments, matching before unseen object instances and person re-identification across spatially disjoint cameras. In the latter two benchmarks we clearly outperform the state-of-the-art.",
"",
"Pose variation remains one of the major factors that adversely affect the accuracy of person re-identification. Such variation is not arbitrary as body parts (e.g. head, torso, legs) have relative stable spatial distribution. Breaking down the variability of global appearance regarding the spatial distribution potentially benefits the person matching. We therefore learn a novel similarity function, which consists of multiple sub-similarity measurements with each taking in charge of a subregion. In particular, we take advantage of the recently proposed polynomial feature map to describe the matching within each subregion, and inject all the feature maps into a unified framework. The framework not only outputs similarity measurements for different regions, but also makes a better consistency among them. Our framework can collaborate local similarities as well as global similarity to exploit their complementary strength. It is flexible to incorporate multiple visual cues to further elevate the performance. In experiments, we analyze the effectiveness of the major components. The results on four datasets show significant and consistent improvements over the state-of-the-art methods.",
"",
"",
"Re-identification of individuals across camera networks with limited or no overlapping fields of view remains challenging in spite of significant research efforts. In this paper, we propose the use, and extensively evaluate the performance, of four alternatives for re-ID classification: regularized Pairwise Constrained Component Analysis, kernel Local Fisher Discriminant Analysis, Marginal Fisher Analysis and a ranking ensemble voting scheme, used in conjunction with different sizes of sets of histogram-based features and linear, χ 2 and RBF-χ 2 kernels. Comparisons against the state-of-art show significant improvements in performance measured both in terms of Cumulative Match Characteristic curves (CMC) and Proportion of Uncertainty Removed (PUR) scores on the challenging VIPeR, iLIDS, CAVIAR and 3DPeS datasets.",
"The accuracy of k-nearest neighbor (kNN) classification depends significantly on the metric used to compute distances between different examples. In this paper, we show how to learn a Mahalanobis distance metric for kNN classification from labeled examples. The Mahalanobis metric can equivalently be viewed as a global linear transformation of the input space that precedes kNN classification using Euclidean distances. In our approach, the metric is trained with the goal that the k-nearest neighbors always belong to the same class while examples from different classes are separated by a large margin. As in support vector machines (SVMs), the margin criterion leads to a convex optimization based on the hinge loss. Unlike learning in SVMs, however, our approach requires no modification or extension for problems in multiway (as opposed to binary) classification. In our framework, the Mahalanobis distance metric is obtained as the solution to a semidefinite program. On several data sets of varying size and difficulty, we find that metrics trained in this way lead to significant improvements in kNN classification. Sometimes these results can be further improved by clustering the training examples and learning an individual metric within each cluster. We show how to learn and combine these local metrics in a globally integrated manner.",
"",
"Person re-identification is an important technique towards automatic search of a person's presence in a surveillance video. Two fundamental problems are critical for person re-identification, feature representation and metric learning. An effective feature representation should be robust to illumination and viewpoint changes, and a discriminant metric should be learned to match various person images. In this paper, we propose an effective feature representation called Local Maximal Occurrence (LOMO), and a subspace and metric learning method called Cross-view Quadratic Discriminant Analysis (XQDA). The LOMO feature analyzes the horizontal occurrence of local features, and maximizes the occurrence to make a stable representation against viewpoint changes. Besides, to handle illumination variations, we apply the Retinex transform and a scale invariant texture operator. To learn a discriminant metric, we propose to learn a discriminant low dimensional subspace by cross-view quadratic discriminant analysis, and simultaneously, a QDA metric is learned on the derived subspace. We also present a practical computation method for XQDA, as well as its regularization. Experiments on four challenging person re-identification databases, VIPeR, QMUL GRID, CUHK Campus, and CUHK03, show that the proposed method improves the state-of-the-art rank-1 identification rates by 2.2 , 4.88 , 28.91 , and 31.55 on the four databases, respectively."
]
} |
1707.07791 | 2736538122 | Learning the distance metric between pairs of examples is of great importance for visual recognition, especially for person re-identification (Re-Id). Recently, the contrastive and triplet loss are proposed to enhance the discriminative power of the deeply learned features, and have achieved remarkable success. As can be seen, either the contrastive or triplet loss is just one special case of the Euclidean distance relationships among these training samples. Therefore, we propose a structured graph Laplacian embedding algorithm, which can formulate all these structured distance relationships into the graph Laplacian form. The proposed method can take full advantages of the structured distance relationships among these training samples, with the constructed complete graph. Besides, this formulation makes our method easy-to-implement and super-effective. When embedding the proposed algorithm with the softmax loss for the CNN training, our method can obtain much more robust and discriminative deep features with inter-personal dispersion and intra-personal compactness, which is essential to person Re-Id. We illustrate the effectiveness of our proposed method on top of three popular networks, namely AlexNet, DGDNet and ResNet50, on recent four widely used Re-Id benchmark datasets. Our proposed method achieves state-of-the-art performances. | Recently, Xiao @cite_54 formulated the person Re-Id problem as a conventional classification problem. They adopted the softmax loss as the identification term, to train one reduced GoogleNet model and achieved superior performance on most person Re-Id benchmark datasets. This indicates that, using the identification loss on top of the deeper network architectures with large datasets, can help achieve good performances. Thus, there are many works aiming to combine softmax loss with other contrastive or verification loss to further improve the Re-Id performances @cite_8 @cite_50 . Combining the two losses aims to exploit both strengths: the classification loss pulls different classes apart and the verification or contrastive loss can reduce the intra-class variations. All of the recently proposed joint learning methods for person Re-Id @cite_50 @cite_32 @cite_56 have two or three network branches, in order to obtain complementary properties. However, as our proposed structured graph Laplacian embedding enjoys the same requirement as the softmax loss and needs no extra recombination of the training data, thus the joint learning only needs one network branch, which makes the CNN training very easy and more efficient. | {
"cite_N": [
"@cite_8",
"@cite_54",
"@cite_32",
"@cite_56",
"@cite_50"
],
"mid": [
"2556388400",
"2952941004",
"2486219732",
"2953350812",
"2549957142"
],
"abstract": [
"Person Re-Identification (re-id) is a challenging task in computer vision, especially when there are limited training data from multiple camera views. In this paper, we pro- pose a deep learning based person re-identification method by transferring knowledge of mid-level attribute features and high-level classification features. Building on the idea that identity classification, attribute recognition and re- identification share the same mid-level semantic representations, they can be trained sequentially by fine-tuning one based on another. In our framework, we train identity classification and attribute recognition tasks from deep Convolutional Neural Network (dCNN) to learn person information. The information can be transferred to the person re-id task and improves its accuracy by a large margin. Further- more, a Long Short Term Memory(LSTM) based Recurrent Neural Network (RNN) component is extended by a spacial gate. This component is used in the re-id model to pay attention to certain spacial parts in each recurrent unit. Experimental results show that our method achieves 78.3 of rank-1 recognition accuracy on the CUHK03 benchmark.",
"Learning generic and robust feature representations with data from multiple domains for the same problem is of great value, especially for the problems that have multiple datasets but none of them are large enough to provide abundant data variations. In this work, we present a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks (CNNs). When training a CNN with data from all the domains, some neurons learn representations shared across several domains, while some others are effective only for a specific one. Based on this important observation, we propose a Domain Guided Dropout algorithm to improve the feature learning procedure. Experiments show the effectiveness of our pipeline and the proposed algorithm. Our methods on the person re-identification problem outperform state-of-the-art methods on multiple datasets by large margins.",
"Person re-identification (ReID) focuses on identifying people across different scenes in video surveillance, which is usually formulated as a binary classification task or a ranking task in current person ReID approaches. In this paper, we take both tasks into account and propose a multi-task deep network (MTDnet) that makes use of their own advantages and jointly optimize the two tasks simultaneously for person ReID. To the best of our knowledge, we are the first to integrate both tasks in one network to solve the person ReID. We show that our proposed architecture significantly boosts the performance. Furthermore, deep architecture in general requires a sufficient dataset for training, which is usually not met in person ReID. To cope with this situation, we further extend the MTDnet and propose a cross-domain architecture that is capable of using an auxiliary set to assist training on small target sets. In the experiments, our approach outperforms most of existing person ReID algorithms on representative datasets including CUHK03, CUHK01, VIPeR, iLIDS and PRID2011, which clearly demonstrates the effectiveness of the proposed approach.",
"Person re-identification (Re-ID) poses a unique challenge to deep learning: how to learn a deep model with millions of parameters on a small training set of few or no labels. In this paper, a number of deep transfer learning models are proposed to address the data sparsity problem. First, a deep network architecture is designed which differs from existing deep Re-ID models in that (a) it is more suitable for transferring representations learned from large image classification datasets, and (b) classification loss and verification loss are combined, each of which adopts a different dropout strategy. Second, a two-stepped fine-tuning strategy is developed to transfer knowledge from auxiliary datasets. Third, given an unlabelled Re-ID dataset, a novel unsupervised deep transfer learning model is developed based on co-training. The proposed models outperform the state-of-the-art deep Re-ID models by large margins: we achieve Rank-1 accuracy of 85.4 , 83.7 and 56.3 on CUHK03, Market1501, and VIPeR respectively, whilst on VIPeR, our unsupervised model (45.1 ) beats most supervised models.",
"In this article, we revisit two popular convolutional neural networks in person re-identification (re-ID): verification and identification models. The two models have their respective advantages and limitations due to different loss functions. Here, we shed light on how to combine the two models to learn more discriminative pedestrian descriptors. Specifically, we propose a Siamese network that simultaneously computes the identification loss and verification loss. Given a pair of training images, the network predicts the identities of the two input images and whether they belong to the same identity. Our network learns a discriminative embedding and a similarity measurement at the same time, thus taking full usage of the re-ID annotations. Our method can be easily applied on different pretrained networks. Albeit simple, the learned embedding improves the state-of-the-art performance on two public person re-ID benchmarks. Further, we show that our architecture can also be applied to image retrieval. The code is available at https: github.com layumi 2016_person_re-ID."
]
} |
1707.07819 | 2737422504 | In this paper, we address the task of detecting semantic parts on partially occluded objects. We consider a scenario where the model is trained using non-occluded images but tested on occluded images. The motivation is that there are infinite number of occlusion patterns in real world, which cannot be fully covered in the training data. So the models should be inherently robust and adaptive to occlusions instead of fitting learning the occlusion patterns in the training data. Our approach detects semantic parts by accumulating the confidence of local visual cues. Specifically, the method uses a simple voting method, based on log-likelihood ratio tests and spatial constraints, to combine the evidence of local cues. These cues are called visual concepts, which are derived by clustering the internal states of deep networks. We evaluate our voting scheme on the VehicleSemanticPart dataset with dense part annotations. We randomly place two, three or four irrelevant objects onto the target object to generate testing images with various occlusions. Experiments show that our algorithm outperforms several competitors in semantic part detection when occlusions are present. | Object detection is a fundamental task in computer vision. As the fast development of deep neural networks @cite_22 @cite_8 @cite_15 , this field has been recently dominated by one type of pipeline @cite_23 @cite_16 , which first generates a set of object proposals @cite_24 @cite_5 , and then predicts the object class of each proposal. This framework has significantly outperformed conventional approaches, which are based on handcrafted features @cite_29 and deformable part models @cite_32 @cite_1 . | {
"cite_N": [
"@cite_22",
"@cite_8",
"@cite_29",
"@cite_1",
"@cite_32",
"@cite_24",
"@cite_23",
"@cite_5",
"@cite_15",
"@cite_16"
],
"mid": [
"",
"1686810756",
"2161969291",
"166750225",
"2168356304",
"2066624635",
"",
"2088049833",
"2194775991",
"2613718673"
],
"abstract": [
"",
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.",
"Deformable part-based models [1, 2] achieve state-of-the-art performance for object detection, but rely on heuristic initialization during training due to the optimization of non-convex cost function. This paper investigates limitations of such an initialization and extends earlier methods using additional supervision. We explore strong supervision in terms of annotated object parts and use it to (i) improve model initialization, (ii) optimize model structure, and (iii) handle partial occlusions. Our method is able to deal with sub-optimal and incomplete annotations of object parts and is shown to benefit from semi-supervised learning setups where part-level annotation is provided for a fraction of positive examples only. Experimental results are reported for the detection of six animal classes in PASCAL VOC 2007 and 2010 datasets. We demonstrate significant improvements in detection performance compared to the LSVM [1] and the Poselet [3] object detectors.",
"We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function.",
"We present a generic objectness measure, quantifying how likely it is for an image window to contain an object of any class. We explicitly train it to distinguish objects with a well-defined boundary in space, such as cows and telephones, from amorphous background elements, such as grass and road. The measure combines in a Bayesian framework several image cues measuring characteristics of objects, such as appearing different from their surroundings and having a closed boundary. These include an innovative cue to measure the closed boundary characteristic. In experiments on the challenging PASCAL VOC 07 dataset, we show this new cue to outperform a state-of-the-art saliency measure, and the combined objectness measure to perform better than any cue alone. We also compare to interest point operators, a HOG detector, and three recent works aiming at automatic object segmentation. Finally, we present two applications of objectness. In the first, we sample a small numberof windows according to their objectness probability and give an algorithm to employ them as location priors for modern class-specific object detectors. As we show experimentally, this greatly reduces the number of windows evaluated by the expensive class-specific model. In the second application, we use objectness as a complementary score in addition to the class-specific model, which leads to fewer false positives. As shown in several recent papers, objectness can act as a valuable focus of attention mechanism in many other applications operating on image windows, including weakly supervised learning of object categories, unsupervised pixelwise segmentation, and object tracking in video. Computing objectness is very efficient and takes only about 4 sec. per image.",
"",
"This paper addresses the problem of generating possible object locations for use in object recognition. We introduce selective search which combines the strength of both an exhaustive search and segmentation. Like segmentation, we use the image structure to guide our sampling process. Like exhaustive search, we aim to capture all possible object locations. Instead of a single technique to generate possible object locations, we diversify our search and use a variety of complementary image partitionings to deal with as many image conditions as possible. Our selective search results in a small set of data-driven, class-independent, high quality locations, yielding 99 recall and a Mean Average Best Overlap of 0.879 at 10,097 locations. The reduced number of locations compared to an exhaustive search enables the use of stronger machine learning techniques and stronger appearance models for object recognition. In this paper we show that our selective search enables the use of the powerful Bag-of-Words model for recognition. The selective search software is made publicly available (Software: http: disi.unitn.it uijlings SelectiveSearch.html ).",
"Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57 error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28 relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [7] and Fast R-CNN [5] have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolutional features. For the very deep VGG-16 model [19], our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2 mAP) and 2012 (70.4 mAP) using 300 proposals per image. Code is available at https: github.com ShaoqingRen faster_rcnn."
]
} |
1707.07819 | 2737422504 | In this paper, we address the task of detecting semantic parts on partially occluded objects. We consider a scenario where the model is trained using non-occluded images but tested on occluded images. The motivation is that there are infinite number of occlusion patterns in real world, which cannot be fully covered in the training data. So the models should be inherently robust and adaptive to occlusions instead of fitting learning the occlusion patterns in the training data. Our approach detects semantic parts by accumulating the confidence of local visual cues. Specifically, the method uses a simple voting method, based on log-likelihood ratio tests and spatial constraints, to combine the evidence of local cues. These cues are called visual concepts, which are derived by clustering the internal states of deep networks. We evaluate our voting scheme on the VehicleSemanticPart dataset with dense part annotations. We randomly place two, three or four irrelevant objects onto the target object to generate testing images with various occlusions. Experiments show that our algorithm outperforms several competitors in semantic part detection when occlusions are present. | Visual concepts @cite_18 are obtained by clustering the intermediate neural responses of deep networks. It is shown that on rigid objects, image patches corresponding to the same visual concept are often visually similar, and that visual concepts are fairly effective in detecting keypoints in the PASCAL3D+ dataset @cite_14 . Our studies are built on previous studies which showed that filters in deep networks often exhibited preferences in stimuli @cite_31 . | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_31"
],
"mid": [
"2552264258",
"1991264156",
""
],
"abstract": [
"We address the key question of how object part representations can be found from the internal states of CNNs that are trained for high-level tasks, such as object classification. This work provides a new unsupervised method to learn semantic parts and gives new understanding of the internal representations of CNNs. Our technique is based on the hypothesis that semantic parts are represented by populations of neurons rather than by single filters. We propose a clustering technique to extract part representations, which we call Visual Concepts. We show that visual concepts are semantically coherent in that they represent semantic parts, and visually coherent in that corresponding image patches appear very similar. Also, visual concepts provide full spatial coverage of the parts of an object, rather than a few sparse parts as is typically found in keypoint annotations. Furthermore, We treat single visual concept as part detector and evaluate it for keypoint detection using the PASCAL3D+ dataset and for part detection using our newly annotated ImageNetPart dataset. The experiments demonstrate that visual concepts can be used to detect parts. We also show that some visual concepts respond to several semantic parts, provided these parts are visually similar. Thus visual concepts have the essential properties: semantic meaning and detection capability. Note that our ImageNetPart dataset gives rich part annotations which cover the whole object, making it useful for other part-related applications.",
"3D object detection and pose estimation methods have become popular in recent years since they can handle ambiguities in 2D images and also provide a richer description for objects compared to 2D object detectors. However, most of the datasets for 3D recognition are limited to a small amount of images per category or are captured in controlled environments. In this paper, we contribute PASCAL3D+ dataset, which is a novel and challenging dataset for 3D object detection and pose estimation. PASCAL3D+ augments 12 rigid categories of the PASCAL VOC 2012 [4] with 3D annotations. Furthermore, more images are added for each category from ImageNet [3]. PASCAL3D+ images exhibit much more variability compared to the existing 3D datasets, and on average there are more than 3,000 object instances per category. We believe this dataset will provide a rich testbed to study 3D detection and pose estimation and will help to significantly push forward research in this area. We provide the results of variations of DPM [6] on our new dataset for object detection and viewpoint estimation in different scenarios, which can be used as baselines for the community. Our benchmark is available online at http: cvgl.stanford.edu projects pascal3d",
""
]
} |
1707.07819 | 2737422504 | In this paper, we address the task of detecting semantic parts on partially occluded objects. We consider a scenario where the model is trained using non-occluded images but tested on occluded images. The motivation is that there are infinite number of occlusion patterns in real world, which cannot be fully covered in the training data. So the models should be inherently robust and adaptive to occlusions instead of fitting learning the occlusion patterns in the training data. Our approach detects semantic parts by accumulating the confidence of local visual cues. Specifically, the method uses a simple voting method, based on log-likelihood ratio tests and spatial constraints, to combine the evidence of local cues. These cues are called visual concepts, which are derived by clustering the internal states of deep networks. We evaluate our voting scheme on the VehicleSemanticPart dataset with dense part annotations. We randomly place two, three or four irrelevant objects onto the target object to generate testing images with various occlusions. Experiments show that our algorithm outperforms several competitors in semantic part detection when occlusions are present. | Occlusion is a common difficulty in object detection @cite_4 or segmentation @cite_13 . @cite_6 used And-or-Graph (AOG) to model the occlusion patterns for car detection. Part-based models is very useful for detecting partially occluded objects @cite_11 @cite_6 . Our method is also part-based but applied to detecting semantic parts. | {
"cite_N": [
"@cite_13",
"@cite_4",
"@cite_6",
"@cite_11"
],
"mid": [
"2344939607",
"",
"225438136",
"2952435733"
],
"abstract": [
"We consider the problem of amodal instance segmentation, the objective of which is to predict the region encompassing both visible and occluded parts of each object. Thus far, the lack of publicly available amodal segmentation annotations has stymied the development of amodal segmentation methods. In this paper, we sidestep this issue by relying solely on standard modal instance segmentation annotations to train our model. The result is a new method for amodal instance segmentation, which represents the first such method to the best of our knowledge. We demonstrate the proposed method’s effectiveness both qualitatively and quantitatively.",
"",
"This paper presents a method of learning reconfigurable hierarchical And-Or models to integrate context and occlusion for car detection. The And-Or model represents the regularities of car-to-car context and occlusion patterns at three levels: (i) layouts of spatially-coupled N cars, (ii) single cars with different viewpoint-occlusion configurations, and (iii) a small number of parts. The learning process consists of two stages. We first learn the structure of the And-Or model with three components: (a) mining N-car contextual patterns based on layouts of annotated single car bounding boxes, (b) mining the occlusion configurations based on the overlapping statistics between single cars, and (c) learning visible parts based on car 3D CAD simulation or heuristically mining latent car parts. The And-Or model is organized into a directed and acyclic graph which leads to the Dynamic Programming algorithm in inference. In the second stage, we jointly train the model parameters (for appearance, deformation and bias) using Weak-Label Structural SVM. In experiments, we test our model on four car datasets: the KITTI dataset [11], the street parking dataset [19], the PASCAL VOC2007 car dataset [7], and a self-collected parking lot dataset. We compare with state-of-the-art variants of deformable part-based models and other methods. Our model obtains significant improvement consistently on the four datasets.",
"This paper presents an approach to parsing humans when there is significant occlusion. We model humans using a graphical model which has a tree structure building on recent work [32, 6] and exploit the connectivity prior that, even in presence of occlusion, the visible nodes form a connected subtree of the graphical model. We call each connected subtree a flexible composition of object parts. This involves a novel method for learning occlusion cues. During inference we need to search over a mixture of different flexible models. By exploiting part sharing, we show that this inference can be done extremely efficiently requiring only twice as many computations as searching for the entire object (i.e., not modeling occlusion). We evaluate our model on the standard benchmarked \"We Are Family\" Stickmen dataset and obtain significant performance improvements over the best alternative algorithms."
]
} |
1707.07819 | 2737422504 | In this paper, we address the task of detecting semantic parts on partially occluded objects. We consider a scenario where the model is trained using non-occluded images but tested on occluded images. The motivation is that there are infinite number of occlusion patterns in real world, which cannot be fully covered in the training data. So the models should be inherently robust and adaptive to occlusions instead of fitting learning the occlusion patterns in the training data. Our approach detects semantic parts by accumulating the confidence of local visual cues. Specifically, the method uses a simple voting method, based on log-likelihood ratio tests and spatial constraints, to combine the evidence of local cues. These cues are called visual concepts, which are derived by clustering the internal states of deep networks. We evaluate our voting scheme on the VehicleSemanticPart dataset with dense part annotations. We randomly place two, three or four irrelevant objects onto the target object to generate testing images with various occlusions. Experiments show that our algorithm outperforms several competitors in semantic part detection when occlusions are present. | Our voting method is similar to @cite_21 @cite_28 @cite_19 . But, we incorporate log-likelihood ratio test @cite_25 and spatial constraint @cite_3 as key components into the voting method, which is new. Also we address a new problem of detecting semantic parts under occlusions, where the training images and testing images are quite different. | {
"cite_N": [
"@cite_28",
"@cite_21",
"@cite_3",
"@cite_19",
"@cite_25"
],
"mid": [
"2106255337",
"2164877691",
"2103497972",
"2543332268",
"193120421"
],
"abstract": [
"We present a discriminative Hough transform based object detector where each local part casts a weighted vote for the possible locations of the object center. We show that the weights can be learned in a max-margin framework which directly optimizes the classification performance. The discriminative training takes into account both the codebook appearance and the spatial distribution of its position with respect to the object center to derive its importance. On various datasets we show that the discriminative training improves the Hough detector. Combined with a verification step using a SVM based classifier, our approach achieves a detection rate of 91.9 at 0.3 false positives per image on the ETHZ shape dataset, a significant improvement over the state of the art, while running the verification step on at least an order of magnitude fewer windows than in a sliding window approach.",
"We present a method for object categorization in real-world scenes. Following a common consensus in the field, we do not assume that a figure- ground segmentation is available prior to recognition. However, in contrast to most standard approaches for object class recognition, our approach automati- cally segments the object as a result of the categorization. This combination of recognition and segmentation into one process is made pos- sible by our use of an Implicit Shape Model, which integrates both into a common probabilistic framework. In addition to the recognition and segmentation result, it also generates a per-pixel confidence measure specifying the area that supports a hypothesis and how much it can be trusted. We use this confidence to derive a nat- ural extension of the approach to handle multiple objects in a scene and resolve ambiguities between overlapping hypotheses with a novel MDL-based criterion. In addition, we present an extensive evaluation of our method on a standard dataset for car detection and compare its performance to existing methods from the literature. Our results show that the proposed method significantly outper- forms previously published methods while needing one order of magnitude less training examples. Finally, we present results for articulated objects, which show that the proposed method can categorize and segment unfamiliar objects in differ- ent articulations and with widely varying texture patterns, even under significant partial occlusion.",
"With contributions from Tomas LozanoPerez and Daniel P. Huttenlocher.An intelligent system must know \"what \"the objects are and \"where \"they are in its environment. Examples of this ubiquitous problem in computer vision arise in tasks involving hand-eye coordination (such as assembling or sorting), inspection tasks, gauging operations, and in navigation and localization of mobile robots. This book describes an extended series of experiments into the role of geometry in the critical area of object recognition. It provides precise definitions of the recognition and localization problems, describes the methods used to address them, analyzes the solutions to these problems, and addresses the implications of this analysis.The solution to problems of object recognition are of fundamental importance in many real applications and versions of the techniques described here are already being used in industrial settings. Although a number of questions remain to be solved, the authors provide a valuable framework for understanding both the strengths and limitations of using object shape to guide recognition.W. Eric L. Grimson is Matsushita Associate Professor in the Department of Electrical Engineering and Computer Science at MIT.Contents: Introduction. Recognition as a Search Problem. Searching for Correspondences. Two-Dimensional Constraints. Three-Dimensional Constraints. Verifying Hypotheses. Controlling the Search Explosion. Selecting Subspaces of the Search Space. Empirical Testing. The Combinatorics of the Matching Process. The Combinatorics of Hough Transforms. The Combinatorics of Verification. The Combinatorics of Indexing. Evaluating the Methods. Recognition from Libraries. Parameterized Objects. The Role of Grouping. Sensing Strategies. Applications. The Next Steps.",
"This paper present a part-based approach for detecting objects with large variation of appearance. We extract local image patches as local features both from the object and from the background in training images to learn an object part model discriminatively. Our object part model discriminates the local features whether they are an object part or not. Based on the discrimination results, each local feature casts probabilistic votes for the object location and size which are learned from the training images. Our object part model also requires regression performance for predicting the object location and size through the voting procedure. We build such an object part model with an ensemble of randomized trees trained by splitting each tree node so as to reduce the entropy of class label distribution and the variance of object location and size. Experimental results on hand detection with large pose variation show that our approach outperforms conventional generalized Hough transform. We verified the performance on a public dataset of side-view cars.",
"From the Publisher: Two important subproblems of computer vision are the detection and recognition of 2D objects in gray-level images. This book discusses the construction and training of models, computational approaches to efficient implementation, and parallel implementations in biologically plausible neural network architectures. The approach is based on statistical modeling and estimation, with an emphasis on simplicity, transparency, and computational efficiency. The book describes a range of deformable template models, from coarse sparse models involving discrete, fast computations to more finely detailed models based on continuum formulations, involving intensive optimization. Each model is defined in terms of a subset of points on a reference grid (the template), a set of admissible instantiations of these points (deformations), and a statistical model for the data given a particular instantiation of the object present in the image. A recurring theme is a coarse to fine approach to the solution of vision problems. The book provides detailed descriptions of the algorithms used as well as the code, and the software and data sets are available on the Web."
]
} |
1707.07413 | 2739427748 | In this work, we perform an empirical comparison among the CTC, RNN-Transducer, and attention-based Seq2Seq models for end-to-end speech recognition. We show that, without any language model, Seq2Seq and RNN-Transducer models both outperform the best reported CTC models with a language model, on the popular Hub5'00 benchmark. On our internal diverse dataset, these trends continue - RNNTransducer models rescored with a language model after beam search outperform our best CTC models. These results simplify the speech recognition pipeline so that decoding can now be expressed purely as neural network operations. We also study how the choice of encoder architecture affects the performance of the three models - when all encoder layers are forward only, and when encoders downsample the input representation aggressively. | Segmental RNNs @cite_13 provide another alternative way to model the ASR task. Segmental RNNs model @math using a zeroth-order CRF. While global normalization help address the label bias issues in CTC, we believe that the bigger issue is still the conditional independence assumptions made by both CTC and Segmental RNNs. | {
"cite_N": [
"@cite_13"
],
"mid": [
"2291022022"
],
"abstract": [
"We study the segmental recurrent neural network for end-to-end acoustic modelling. This model connects the segmental conditional random field (CRF) with a recurrent neural network (RNN) used for feature extraction. Compared to most previous CRF-based acoustic models, it does not rely on an external system to provide features or segmentation boundaries. Instead, this model marginalises out all the possible segmentations, and features are extracted from the RNN trained together with the segmental CRF. In essence, this model is self-contained and can be trained end-to-end. In this paper, we discuss practical training and decoding issues as well as the method to speed up the training in the context of speech recognition. We performed experiments on the TIMIT dataset. We achieved 17.3 phone error rate (PER) from the first-pass decoding --- the best reported result using CRFs, despite the fact that we only used a zeroth-order CRF and without using any language model."
]
} |
1707.07343 | 2739371250 | We present a sequential model for temporal relation classification between intra-sentence events. The key observation is that the overall syntactic structure and compositional meanings of the multi-word context between events are important for distinguishing among fine-grained temporal relations. Specifically, our approach first extracts a sequence of context words that indicates the temporal relation between two events, which well align with the dependency path between two event mentions. The context word sequence, together with a parts-of-speech tag sequence and a dependency relation sequence that are generated corresponding to the word sequence, are then provided as input to bidirectional recurrent neural network (LSTM) models. The neural nets learn compositional syntactic and semantic representations of contexts surrounding the two events and predict the temporal relation between them. Evaluation of the proposed approach on TimeBank corpus shows that sequential modeling is capable of accurately recognizing temporal relations between events, which outperforms a neural net model using various discrete features as input that imitates previous feature based models. | The following works mostly expanded the feature sets @cite_11 @cite_10 @cite_4 @cite_16 @cite_15 @cite_18 @cite_12 . Specifically, used direct dependency path between event pairs to capture syntactic context. used 3-grams of paths between two event mentions in a dependency tree as features instead of full paths as those are too sparse. We found that modeling the entire path as one sequence provides greater compositional evidence on the temporal relation. In addition, modifiers attached to the words in a path with specific dependency relations like nmod:tmod are also informative. | {
"cite_N": [
"@cite_18",
"@cite_4",
"@cite_15",
"@cite_16",
"@cite_10",
"@cite_12",
"@cite_11"
],
"mid": [
"2109308402",
"1571872475",
"2115678331",
"2251873637",
"",
"2060381885",
""
],
"abstract": [
"This paper describes a complete event time ordering system that annotates raw text with events, times, and the ordering relations between them at the SemEval-2013 Task 1. Task 1 is a unique challenge because it starts from raw text, rather than pre-annotated text with known events and times. A working system first identifies events and times, then identifies which events and times should be ordered, and finally labels the ordering relation between them. We present a split classifier approach that breaks the ordering tasks into smaller decision points. Experiments show that more specialized classifiers perform better than few joint classifiers. The NavyTime system ranked second both overall and in most subtasks like event extraction and relation labeling.",
"We describe the TempEval-3 task which is currently in preparation for the SemEval-2013 evaluation exercise. The aim of TempEval is to advance research on temporal information processing. TempEval-3 follows on from previous TempEval events, incorporating: a three-part task structure covering event, temporal expression and temporal relation extraction; a larger dataset; and single overall task quality scores.",
"We propose a new approach to characterizing the timeline of a text: temporal dependency structures, where all the events of a narrative are linked via partial ordering relations like BEFORE, AFTER, OVERLAP and IDENTITY. We annotate a corpus of children's stories with temporal dependency trees, achieving agreement (Krippendorff's Alpha) of 0.856 on the event words, 0.822 on the links between events, and of 0.700 on the ordering relation labels. We compare two parsing models for temporal dependency structures, and show that a deterministic non-projective dependency parser outperforms a graph-based maximum spanning tree parser, achieving labeled attachment accuracy of 0.647 and labeled tree edit distance of 0.596. Our analysis of the dependency parser errors gives some insights into future research directions.",
"The ClearTK-TimeML submission to TempEval 2013 competed in all English tasks: identifying events, identifying times, and identifying temporal relations. The system is a pipeline of machine-learning models, each with a small set of features from a simple morpho-syntactic annotation pipeline, and where temporal relations are only predicted for a small set of syntactic constructions and relation types. ClearTKTimeML ranked 1 st for temporal relation F1, time extent strict F1 and event tense accuracy.",
"",
"We approached the temporal relation identification tasks of TempEval 2007 as pair-wise classification tasks. We introduced a variety of syntactically and semantically motivated features, including temporal-logic-based features derived from running our Task B system on the Task A and C data. We trained support vector machine models and achieved the second highest accuracies on the tasks: 61 on Task A, 75 on Task B and 54 on Task C.",
""
]
} |
1707.07273 | 2738508272 | We present a novel neural model HyperVec to learn hierarchical embeddings for hypernymy detection and directionality. While previous embeddings have shown limitations on prototypical hypernyms, HyperVec represents an unsupervised measure where embeddings are learned in a specific order and capture the hypernym @math hyponym distributional hierarchy. Moreover, our model is able to generalize over unseen hypernymy pairs, when using only small sets of training data, and by mapping to other languages. Results on benchmark datasets show that HyperVec outperforms both state @math of @math the @math art unsupervised measures and embedding models on hypernymy detection and directionality, and on predicting graded lexical entailment. | A variety of directional measures for unsupervised hypernymy detection @cite_20 @cite_5 @cite_19 @cite_37 @cite_27 all rely on some variation of the : If @math is a semantically narrower term than @math , then a significant number of salient distributional features of @math is expected to be included in the feature vector of @math as well. In addition, proposed the , that hypernyms tend to be less informative than hyponyms, and that they occur in more general contexts than their hyponyms. All of these approaches represent words as vectors in distributional semantic models @cite_23 , relying on the @cite_28 @cite_21 . For evaluation, these directional models use the AP measure to assess the proportion of hypernyms at the top of a score-sorted list. In a different vein, introduced three unsupervised methods drawn from visual properties of images to determine a concept's generality in hypernymy tasks. | {
"cite_N": [
"@cite_37",
"@cite_28",
"@cite_21",
"@cite_19",
"@cite_27",
"@cite_23",
"@cite_5",
"@cite_20"
],
"mid": [
"2131253837",
"",
"",
"2100258351",
"141602984",
"1662133657",
"1966907789",
"2101941521"
],
"abstract": [
"Distributional word similarity is most commonly perceived as a symmetric relation. Yet, directional relations are abundant in lexical semantics and in many Natural Language Processing (NLP) settings that require lexical inference, making symmetric similarity measures less suitable for their identification. This paper investigates the nature of directional (asymmetric) similarity measures that aim to quantify distributional feature inclusion. We identify desired properties of such measures for lexical inference, specify a particular measure based on Average Precision that addresses these properties, and demonstrate the empirical benefit of directional measures for two different NLP datasets.",
"",
"",
"We present the context-theoretic framework, which provides a set of rules for the nature of composition of meaning based on the philosophy of meaning as context. Principally, in the framework the composition of the meaning of words can be represented as multiplication of their representative vectors, where multiplication is distributive with respect to the vector space. We discuss the applicability of the framework to a range of techniques in natural language processing, including subsequence matching, the lexical entailment model of (2005), vector-based representations of taxonomies, statistical parsing and the representation of uncertainty in logical semantics.",
"In this paper we apply existing directional similarity measures to identify hypernyms with a state-of-the-art distributional semantic model. We also propose a new directional measure that achieves the best performance in hypernym identification.",
"Computers understand very little of the meaning of human language. This profoundly limits our ability to give instructions to computers, the ability of computers to explain their actions to us, and the ability of computers to analyse and process text. Vector space models (VSMs) of semantics are beginning to address these limits. This paper surveys the use of VSMs for semantic processing of text. We organize the literature on VSMs according to the structure of the matrix in a VSM. There are currently three broad classes of VSMs, based on term-document, word-context, and pair-pattern matrices, yielding three classes of applications. We survey a broad range of applications in these three categories and we take a detailed look at a specific open source project in each category. Our goal in this survey is to show the breadth of applications of VSMs for semantics, to provide a new perspective on VSMs for those who are already familiar with the area, and to provide pointers into the literature for those who are less familiar with the field.",
"This work investigates the variation in a word's distributionally nearest neighbours with respect to the similarity measure used. We identify one type of variation as being the relative frequency of the neighbour words with respect to the frequency of the target word. We then demonstrate a three-way connection between relative frequency of similar words, a concept of distributional gnerality and the semantic relation of hyponymy. Finally, we consider the impact that this has on one application of distributional similarity methods (judging the compositionality of collocations).",
"We present a general framework for distributional similarity based on the concepts of precision and recall. Different parameter settings within this framework approximate different existing similarity measures as well as many more which have, until now, been unexplored. We show that optimal parameter settings outperform two existing state-of-the-art similarity measures on two evaluation tasks for high and low frequency nouns."
]
} |
1707.07273 | 2738508272 | We present a novel neural model HyperVec to learn hierarchical embeddings for hypernymy detection and directionality. While previous embeddings have shown limitations on prototypical hypernyms, HyperVec represents an unsupervised measure where embeddings are learned in a specific order and capture the hypernym @math hyponym distributional hierarchy. Moreover, our model is able to generalize over unseen hypernymy pairs, when using only small sets of training data, and by mapping to other languages. Results on benchmark datasets show that HyperVec outperforms both state @math of @math the @math art unsupervised measures and embedding models on hypernymy detection and directionality, and on predicting graded lexical entailment. | The studies in this area are based on word embeddings which represent words as low-dimensional and real-valued vectors @cite_39 @cite_40 . Each hypernymy pair is encoded by some combination of the two word vectors, such as concatenation @cite_41 or difference @cite_26 @cite_7 . Hypernymy is distinguished from other relations by using a classification approach, such as SVM or LR. Because word embeddings are trained for similar and symmetric vectors, it is however unclear whether the supervised methods do actually learn the asymmetry in hypernymy @cite_35 . | {
"cite_N": [
"@cite_35",
"@cite_26",
"@cite_7",
"@cite_41",
"@cite_39",
"@cite_40"
],
"mid": [
"",
"1583820951",
"1516501661",
"205765513",
"2950133940",
"2250539671"
],
"abstract": [
"",
"We test the Distributional Inclusion Hypothesis, which states that hypernyms tend to occur in a superset of contexts in which their hyponyms are found. We find that this hypothesis only holds when it is applied to relevant dimensions. We propose a robust supervised approach that achieves accuracies of .84 and .85 on two existing datasets and that can be interpreted as selecting the dimensions that are relevant for distributional inclusion.",
"This work is concerned with distinguishing different semantic relations which exist between distributionally similar words. We compare a novel approach based on training a linear Support Vector Machine on pairs of feature vectors with state-of-the-art methods based on distributional similarity. We show that the new supervised approach does better even when there is minimal information about the target words in the training data, giving a 15 reduction in error rate over unsupervised approaches.",
"We introduce two ways to detect entailment using distributional semantic representations of phrases. Our first experiment shows that the entailment relation between adjective-noun constructions and their head nouns (big cat|= cat), once represented as semantic vector pairs, generalizes to lexical entailment among nouns (dog|= animal). Our second experiment shows that a classifier fed semantic vector pairs can similarly generalize the entailment relation among quantifier phrases (many dogs|= some dogs) to entailment involving unseen quantifiers (all cats|= several cats). Moreover, nominal and quantifier phrase entailment appears to be cued by different distributional correlates, as predicted by the type-based view of entailment in formal semantics.",
"The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling. An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of \"Canada\" and \"Air\" cannot be easily combined to obtain \"Air Canada\". Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible.",
"Recent methods for learning vector space representations of words have succeeded in capturing fine-grained semantic and syntactic regularities using vector arithmetic, but the origin of these regularities has remained opaque. We analyze and make explicit the model properties needed for such regularities to emerge in word vectors. The result is a new global logbilinear regression model that combines the advantages of the two major model families in the literature: global matrix factorization and local context window methods. Our model efficiently leverages statistical information by training only on the nonzero elements in a word-word cooccurrence matrix, rather than on the entire sparse matrix or on individual context windows in a large corpus. The model produces a vector space with meaningful substructure, as evidenced by its performance of 75 on a recent word analogy task. It also outperforms related models on similarity tasks and named entity recognition."
]
} |
1707.07273 | 2738508272 | We present a novel neural model HyperVec to learn hierarchical embeddings for hypernymy detection and directionality. While previous embeddings have shown limitations on prototypical hypernyms, HyperVec represents an unsupervised measure where embeddings are learned in a specific order and capture the hypernym @math hyponym distributional hierarchy. Moreover, our model is able to generalize over unseen hypernymy pairs, when using only small sets of training data, and by mapping to other languages. Results on benchmark datasets show that HyperVec outperforms both state @math of @math the @math art unsupervised measures and embedding models on hypernymy detection and directionality, and on predicting graded lexical entailment. | These approaches are closest to our work. proposed a dynamic distance-margin model to learn term embeddings that capture properties of hypernymy. The neural model is trained on the taxonomic relation data which is pre-extracted. The resulting term embeddings are fed to an SVM classifier to predict hypernymy. However, this model only learns term pairs without considering their contexts, leading to a lack of generalization for term embeddings. introduced a dynamic weighting neural network to learn term embeddings that encode information about hypernymy and also about their contexts, considering all words between a hypernym and its hyponym in a sentence. The proposed model is trained on a set of hypernym relations extracted from WordNet @cite_33 . The embeddings are applied as features to detect hypernymy, using an SVM classifier. handles the drawback of the approach by , considering the contextual information between two terms; however the method still is not able to determine the directionality of a hypernym pair. proposed a method to encode order into learned distributed representations, to explicitly model partial order structure of the visual-semantic hierarchy or the hierarchy of hypernymy in WordNet. The resulting vectors are used to predict the transitive hypernym relations in WordNet. | {
"cite_N": [
"@cite_33"
],
"mid": [
"2081580037"
],
"abstract": [
"Because meaningful sentences are composed of meaningful words, any system that hopes to process natural languages as people do must have information about words and their meanings. This information is traditionally provided through dictionaries, and machine-readable dictionaries are now widely available. But dictionary entries evolved for the convenience of human readers, not for machines. WordNet 1 provides a more effective combination of traditional lexicographic information and modern computing. WordNet is an online lexical database designed for use under program control. English nouns, verbs, adjectives, and adverbs are organized into sets of synonyms, each representing a lexicalized concept. Semantic relations link the synonym sets [4]."
]
} |
1707.07344 | 2738613378 | We introduce a novel iterative approach for event coreference resolution that gradually builds event clusters by exploiting inter-dependencies among event mentions within the same chain as well as across event chains. Among event mentions in the same chain, we distinguish within- and cross-document event coreference links by using two distinct pairwise classifiers, trained separately to capture differences in feature distributions of within- and cross-document event clusters. Our event coreference approach alternates between WD and CD clustering and combines arguments from both event clusters after every merge, continuing till no more merge can be made. And then it performs further merging between event chains that are both closely related to a set of other chains of events. Experiments on the ECB+ corpus show that our model outperforms state-of-the-art methods in joint task of WD and CD event coreference resolution. | Different approaches, focusing on either of WD or CD coreference chains, have been proposed for event coreference resolution. Works specific to WD event coreference includes pairwise classifiers @cite_12 @cite_28 graph based clustering method @cite_18 , information propagation @cite_30 , and markov logic networks . As to only CD event coreference, created pairwise classifiers using features indicating granularities of event slots and in another work ( cybulska2015bag ), grouped events based on compatibilities of event contexts. | {
"cite_N": [
"@cite_28",
"@cite_18",
"@cite_30",
"@cite_12"
],
"mid": [
"60009487",
"2251110706",
"",
"2072628044"
],
"abstract": [
"In past years, there has been substantial work on the problem of entity coreference resolution whereas much less attention has been paid to event coreference resolution. Starting with some motivating examples, we formally state the problem of event coreference resolution in the ACE program, present an agglomerative clustering algorithm for the task, explore the feature impact in the event coreference model and compare three evaluation metrics that were previously adopted in entity coreference resolution: MUC F-Measure, B-Cubed F-Measure and ECM F-Measure.",
"In this paper we focus on a new problem of event coreference resolution across television news videos. Based on the observation that the contents from multiple data modalities are complementary, we develop a novel approach to jointly encode effective features from both closed captions and video key frames. Experiment results demonstrate that visual features provided 7.2 absolute F-score gain on stateof-the-art text based event extraction and coreference resolution.",
"",
"Event detection and recognition is a complex task consisting of multiple sub-tasks of varying difficulty. In this paper, we present a simple, modular approach to event extraction that allows us to experiment with a variety of machine learning methods for these sub-tasks, as well as to evaluate the impact on performance these sub-tasks have on the overall task."
]
} |
1707.07432 | 2738277681 | Offline handwritten text line recognition is a hard task that requires both an efficient optical character recognizer and language model. Handwriting recognition state of the art methods are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN) coupled with the use of linguistic knowledge. Most of the proposed approaches in the literature focus on improving one of the two components and use constraint, dedicated to a database lexicon. However, state of the art performance is achieved by combining multiple optical models, and possibly multiple language models with the Recognizer Output Voting Error Reduction (ROVER) framework. Though handwritten line recognition with ROVER has been implemented by combining only few recognizers because training multiple complete recognizers is hard. In this paper we propose a Lexicon Verified ROVER: LV-ROVER, that has a reduce complexity compare to the original one and that can combine hundreds of recognizers without language models. We achieve state of the art for handwritten line text on the RIMES dataset. | Nowadays, almost apart some preliminary works from a2ia that directly process paragraphs using attention-based deep neural network models every handwriting recognition systems take as input pre-segmented text lines. Text lines were then historically segmented into words and then into characters to train optical character recognizer. Today state-of-the art OCR methods directly recognize lines in their whole, using Long Short Term Memory (LSTM) recurrent neural networks (RNN) trained with the Connectionist Temporal Classification . Despite the significant progress achieved with RNN, they cannot achieve state of the art performance alone. Similar to the approaches introduced some decades ago with Hidden Markov Models , there is a need to combine the RNN optical models with lexicons and language models in a lexicon and language model driven decoding scheme to achieve the state of the art recognition performance @cite_5 @cite_4 @cite_0 . | {
"cite_N": [
"@cite_0",
"@cite_5",
"@cite_4"
],
"mid": [
"2573601984",
"1994420166",
"2133552271"
],
"abstract": [
"Multidimensional long short-term memory recurrent neural networks achieve impressive results for handwriting recognition. However, with current CPU-based implementations, their training is very expensive and thus their capacity has so far been limited. We release an efficient GPU-based implementation which greatly reduces training times by processing the input in a diagonal-wise fashion. We use this implementation to explore deeper and wider architectures than previously used for handwriting recognition and show that especially the depth plays an important role. We outperform state of the art results on two databases with a deep multidimensional network.",
"The use of Language Models (LMs) is a very important component in large and open vocabulary recognition systems. This paper presents an open-vocabulary approach for Arabic handwriting recognition. The proposed approach makes use of Arabic word decomposition based on morphological analysis. The vocabulary is a combination of words and sub-words obtained by the decomposition process. Out Of Vocabulary (OOV) words can be recognized by combining different elements from the lexicon. The recognition system is based on Hidden Markov Models (HMMs) with position and context dependent character models. An n-gram LM trained on the decomposed text is used along with the HMMs during the search. The approach is evaluated using two Arabic handwriting datasets. The open vocabulary approach leads to a significant improvement in the system performance. Two different types experiments for two Arabic handwriting recognition tasks are conducted in this work. The proposed approach for open vocabulary allows to have an absolute improvement of up to 1 in the Word Error Rate (WER) for the constrained task and to keep the same performance of the baseline system for the unconstrained one.",
"Since their first inception more than half a century ago, automatic reading systems have evolved substantially, thereby showing impressive performance on machine-printed text. The recognition of handwriting can, however, still be considered an open research problem due to its substantial variation in appearance. With the introduction of Markovian models to the field, a promising modeling and recognition paradigm was established for automatic offline handwriting recognition. However, so far, no standard procedures for building Markov-model-based recognizers could be established though trends toward unified approaches can be identified. It is therefore the goal of this survey to provide a comprehensive overview of the application of Markov models in the research field of offline handwriting recognition, covering both the widely used hidden Markov models and the less complex Markov-chain or n-gram models. First, we will introduce the typical architecture of a Markov-model-based offline handwriting recognition system and make the reader familiar with the essential theoretical concepts behind Markovian models. Then, we will give a thorough review of the solutions proposed in the literature for the open problems how to apply Markov-model-based approaches to automatic offline handwriting recognition."
]
} |
1707.07432 | 2738277681 | Offline handwritten text line recognition is a hard task that requires both an efficient optical character recognizer and language model. Handwriting recognition state of the art methods are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN) coupled with the use of linguistic knowledge. Most of the proposed approaches in the literature focus on improving one of the two components and use constraint, dedicated to a database lexicon. However, state of the art performance is achieved by combining multiple optical models, and possibly multiple language models with the Recognizer Output Voting Error Reduction (ROVER) framework. Though handwritten line recognition with ROVER has been implemented by combining only few recognizers because training multiple complete recognizers is hard. In this paper we propose a Lexicon Verified ROVER: LV-ROVER, that has a reduce complexity compare to the original one and that can combine hundreds of recognizers without language models. We achieve state of the art for handwritten line text on the RIMES dataset. | Although state-of-the art systems now provide interesting performance @cite_6 , the idea of combining several recognizers has emerged, in order to reach better performance. However, combining the output of several recognizers is not trivial, mainly due to the dynamical nature of the models, that may lead to output signals of different sizes. The most popular combination method is the ROVER method that aligns results of many systems (each operating a lexicon and language model driven decoding scheme), and votes for the most probable solution. ROVER has become a popular solution introduced in many methods of the literature or with some modification . We now present the ROVER combination scheme in the next paragraph. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2095095223"
],
"abstract": [
"Abstract : In this paper it is shown that a combination of the classical Lagrange multiplier formalism and the functional equation technique of dynamic programming enables a number of types of variational problems involving the computation and tabulation of functions of M variables to be treated by computing first sequences of functions of K variables, and then sequences of functions of M--K variables, where K may be chosen within the range 1 or = K or = M-1. The choice of K depends upon the process This reduction in the dimensionality of the functions involved is equivalent to an increase in the capability of modern digital computers as far as dynamic programming processes are concerned."
]
} |
1707.07278 | 2737807951 | is one of the core editing principles in Wikipedia, editors being encouraged to provide citations for the added content. For a Wikipedia article, determining the of a citation, i.e. what content is covered by a citation, is important as it helps decide for which content citations are still missing. We are the first to address the problem of determining the in Wikipedia articles. We approach this problem by classifying which textual fragments in an article are covered by a citation. We propose a sequence classification approach where for a paragraph and a citation, we determine the citation span at a fine-grained level. We provide a thorough experimental evaluation and compare our approach against baselines adopted from the scientific domain, where we show improvement for all evaluation metrics. | One of the first attempts to determine the citation span in text @cite_11 was carried out in the context of document retrieval. The citing statements from a document were used as an index to retrieve the document. The citing statements are extracted based on heuristics starting from the citing sentence and are expanded with sentences in a window of + -2 sentences, depending on them containing cue words like . We consider the approach in @cite_11 as a baseline. | {
"cite_N": [
"@cite_11"
],
"mid": [
"2008479441"
],
"abstract": [
"Abstract Documents in computer-readable form can be used to provide information about other documents, i.e. those they cite. To do this efficiently requires procedures for computer recognition of citing statements. This is not easy, especially for multi-sentence citing statements. Computer recognition procedures have been developed which are accurate to the following extent: 73 of the words in statements selected by computer procedures as being citing statements are words which are correctly attributable to the corresponding documents. The retrieval effectiveness of computer-recognized citing statements was tested in the following way. First, for eight retrieval requests in inorganiic chemistry, average recall by search of Chemical Abstracts Service indexing and Chemical Abstracts abstract text words was found to be 50 . Words from citing statements referring to the papers to be retrieved were then added to the index terms and abstract words as additional access points, and searching was repeated. Average recall increased to 70 . Only words from citing statements published within a year of the cited papers were used. The retrieval effect of citing statement words alone (published within a year) without index or abstract terms was the following: average recall was 40 . When just the words of the titles of the cited papers were added to those citing statement words, average recall increased to 50 ."
]
} |
1707.07334 | 2738327010 | We study which property testing and sublinear time algorithms can be transformed into graph streaming algorithms for random order streams. Our main result is that for bounded degree graphs, any property that is constant-query testable in the adjacency list model can be tested with constant space in a single-pass in random order streams. Our result is obtained by estimating the distribution of local neighborhoods of the vertices on a random order graph stream using constant space. We then show that our approach can also be applied to constant time approximation algorithms for bounded degree graphs in the adjacency list model: As an example, we obtain a constant-space single-pass random order streaming algorithms for approximating the size of a maximum matching with additive error @math ( @math is the number of nodes). Our result establishes for the first time that a large class of sublinear algorithms can be simulated in random order streams, while @math space is needed for many graph streaming problems for adversarial orders. | @cite_27 initiated the study of property testing in streaming model, and they gave efficient testers for some properties of a sequence of data items (rather than graphs as we consider here). Bury and Schwiegelshohn @cite_16 gave a lower bound of @math on the space complexity of any algorithm that @math -approximates the size of maximum matching in adversarial streams. @cite_10 showed that in random streams, @math space is necessary to distinguish if a graph is bipartite or @math -far from being bipartite. Previous work has extensively studied streaming graph algorithms in both the insertion-only and dynamic models, see the recent survey @cite_18 . | {
"cite_N": [
"@cite_27",
"@cite_16",
"@cite_10",
"@cite_18"
],
"mid": [
"1558846832",
"1617753329",
"1824295098",
"2016289973"
],
"abstract": [
"We consider the tasks of testing and spot-checking for data streams . These testers and spot-checkers are potentially useful in real-time or near real-time applications that process huge data sets. Crucial aspects of the computational model include the space complexity of the testers and spot-checkers (ideally much lower than the size of the input stream) and the number of passes that the tester or spot-checker must make over the input stream (ideally one, because the original stream may be too large to store for a second pass).",
"This paper presents an algorithm for estimating the weight of a maximum weighted matching by augmenting any estimation routine for the size of an unweighted matching. The algorithm is implementable in any streaming model including dynamic graph streams. We also give the first constant estimation for the maximum matching size in a dynamic graph stream for planar graphs (or any graph with bounded arboricity) using ( O (n^ 4 5 ) ) space which also extends to weighted matching. Using previous results by Kapralov, Khanna, and Sudan (2014) we obtain a polylog(n) approximation for general graphs using polylog(n) space in random order streams, respectively. In addition, we give a space lower bound of Ω(n1 − e) for any randomized algorithm estimating the size of a maximum matching up to a 1 + O(e) factor for adversarial streams.",
"We consider the problem of estimating the value of max cut in a graph in the streaming model of computation. At one extreme, there is a trivial 2-approximation for this problem that uses only O(log n) space, namely, count the number of edges and output half of this value as the estimate for max cut value. On the other extreme, if one allows O(n) space, then a near-optimal solution to the max cut value can be obtained by storing an O(n)-size sparsifier that essentially preserves the max cut. An intriguing question is if poly-logarithmic space suffices to obtain a non-trivial approximation to the max-cut value (that is, beating the factor 2). It was recently shown that the problem of estimating the size of a maximum matching in a graph admits a non-trivial approximation in poly-logarithmic space. Our main result is that any streaming algorithm that breaks the 2-approximation barrier requires [EQUATION] space even if the edges of the input graph are presented in random order. Our result is obtained by exhibiting a distribution over graphs which are either bipartite or 1 2-far from being bipartite, and establishing that [EQUATION] space is necessary to differentiate between these two cases. Thus as a direct corollary we obtain that [EQUATION] space is also necessary to test if a graph is bipartite or 1 2-far from being bipartite. We also show that for any e > 0, any streaming algorithm that obtains a (1 + e)-approximation to the max cut value when edges arrive in adversarial order requires n1−O(e) space, implying that Ω(n) space is necessary to obtain an arbitrarily good approximation to the max cut value.",
"Over the last decade, there has been considerable interest in designing algorithms for processing massive graphs in the data stream model. The original motivation was two-fold: a) in many applications, the dynamic graphs that arise are too large to be stored in the main memory of a single machine and b) considering graph problems yields new insights into the complexity of stream computation. However, the techniques developed in this area are now finding applications in other areas including data structures for dynamic graphs, approximation algorithms, and distributed and parallel computation. We survey the state-of-the-art results; identify general techniques; and highlight some simple algorithms that illustrate basic ideas."
]
} |
1707.07334 | 2738327010 | We study which property testing and sublinear time algorithms can be transformed into graph streaming algorithms for random order streams. Our main result is that for bounded degree graphs, any property that is constant-query testable in the adjacency list model can be tested with constant space in a single-pass in random order streams. Our result is obtained by estimating the distribution of local neighborhoods of the vertices on a random order graph stream using constant space. We then show that our approach can also be applied to constant time approximation algorithms for bounded degree graphs in the adjacency list model: As an example, we obtain a constant-space single-pass random order streaming algorithms for approximating the size of a maximum matching with additive error @math ( @math is the number of nodes). Our result establishes for the first time that a large class of sublinear algorithms can be simulated in random order streams, while @math space is needed for many graph streaming problems for adversarial orders. | In the framework of @math -bounded graph property testing, it is now known that many interesting properties are constant-query testable in the adjacency list model, including @math -edge connectivity, cycle-freeness, subgraph-freeness @cite_11 , @math -vertex connectivity @cite_28 , minor-freeness @cite_2 @cite_5 , matroids related properties @cite_30 @cite_0 , hyperfinite properties @cite_24 , subdivision-freeness @cite_13 . Constant-time approximation algorithms in @math -bounded graphs are known to exist for a number of fundamental optimization problems, including @math -approximating the weight of minimum spanning tree @cite_15 , @math -approximating the size of maximal maximum matching @cite_17 @cite_3 , @math -approximating the minimum vertex cover size @cite_12 @cite_29 @cite_1 , @math -approximating the minimum dominating set size @cite_12 @cite_17 . For @math -bounded minor-free graphs, there are constant-time @math -approximation algorithms for the size of minimum vertex cover, minimum dominating set and maximum independent set @cite_2 . | {
"cite_N": [
"@cite_30",
"@cite_28",
"@cite_29",
"@cite_1",
"@cite_17",
"@cite_3",
"@cite_0",
"@cite_24",
"@cite_2",
"@cite_5",
"@cite_15",
"@cite_13",
"@cite_12",
"@cite_11"
],
"mid": [
"2950966800",
"",
"",
"2116953739",
"2109330224",
"2072015757",
"1997480274",
"2022140055",
"",
"",
"2082227928",
"2902263058",
"1980155175",
"1980816908"
],
"abstract": [
"A graph @math is called @math -full if @math contains a subgraph @math of @math edges such that, for any non-empty @math , @math holds. Here, @math denotes the set of vertices incident to @math . It is known that the family of edge sets of @math -full graphs forms a family of matroid, known as the sparsity matroid of @math . In this paper, we give a constant-time approximation algorithm for the rank of the sparsity matroid of a degree-bounded undirected graph. This leads to a constant-time tester for @math -fullness in the bounded-degree model, (i.e., we can decide with high probability whether an input graph satisfies a property @math or far from @math ). Depending on the values of @math and @math , it can test various properties of a graph such as connectivity, rigidity, and how many spanning trees can be packed. Based on this result, we also propose a constant-time tester for @math -edge-connected-orientability in the bounded-degree model, where an undirected graph @math is called @math -edge-connected-orientable if there exists an orientation @math of @math with a vertex @math such that @math contains @math arc-disjoint dipaths from @math to each vertex @math and @math arc-disjoint dipaths from each vertex @math to @math . A tester is called a one-sided error tester for @math if it always accepts a graph satisfying @math . We show, for @math and (proper) @math , any one-sided error tester for @math -fullness and @math -edge-connected-orientability requires @math queries.",
"",
"",
"We give a nearly optimal sublinear-time algorithm for approximating the size of a minimum vertex cover in a graph G. The algorithm may query the degree deg(v) of any vertex v of its choice, and for each 1 ≤ i ≤ deg(v), it may ask for the ith neighbor of v. Letting VCopt(G) denote the minimum size of vertex cover in G, the algorithm outputs, with high constant success probability, an estimate [EQUATION] such that [EQUATION], where e is a given additive approximation parameter. We refer to such an estimate as a (2, e)-estimate. The query complexity and running time of the algorithm are O([EQUATION] · poly(1 e)), where d denotes the average vertex degree in the graph. The best previously known sublinear algorithm, of (STOC 2009), has query complexity and running time O(d4 e2), where d is the maximum degree in the graph. Given the lower bound of Ω(d) (for constant e) for obtaining such an estimate (with any constant multiplicative factor) due to Parnas and Ron (TCS 2007), our result is nearly optimal. In the case that the graph is dense, that is, the number of edges is Θ(n2), we consider another model, in which the algorithm may ask, for any pair of vertices u and v, whether there is an edge between u and v. We show how to adapt the algorithm that uses neighbor queries to this model and obtain an algorithm that outputs a (2, e)-estimate of the size of a minimum vertex cover whose query complexity and running time are O(n) · poly(1 e).",
"We present a technique for transforming classical approximation algorithms into constant-time algorithms that approximate the size of the optimal solution. Our technique is applicable to a certain subclass of algorithms that compute a solution in a constant number of phases. The technique is based on greedily considering local improvements in random order.The problems amenable to our technique include vertex cover, maximum matching, maximum weight matching, set cover, and minimum dominating set. For example, for maximum matching, we give the first constant-time algorithm that for the class of graphs of degree bounded by d, computes the maximum matching size to within epsivn, for any epsivn > 0, where n is the number of nodes in the graph. The running time of the algorithm is independent of n, and only depends on d and epsiv.",
"We study constant-time approximation algorithms for bounded-degree graphs, which run in time independent of the number of vertices @math . We present an algorithm that decides whether a vertex is contained in a some fixed maximal independent set with expected query complexity @math , where @math is the degree bound. Using this algorithm, we show constant-time approximation algorithms with certain multiplicative error and additive error @math for many other problems, e.g., the maximum matching problem, the minimum vertex cover problem, and the minimum set cover problem, that run exponentially faster than existing algorithms with respect to @math and @math . Our approximation algorithm for the maximum matching problem can be transformed to a two-sided error tester for the property of having a perfect matching. On the contrary, we show that every one-sided error tester for the property requires at least @math queries.",
"For a function @math on a finite set V with f(?)=f(V)=0, a digraph D=(V,A) is called f-connected if it satisfies the f-cut condition, that is, ? D (X)?f(X) for any X⊆V, where ? D (X) is the number of arcs from X to V?X. We show that, for any crossing supermodular function f, the f-connectivity can be tested with a constant number of queries in the general digraph model with average degree bound. As immediate corollaries, we obtain constant-time testers for k-edge-connectivity, rooted-(k,l)-edge-connectivity, and the property of having k arc-disjoint arborescences. We also give a corresponding result for the undirected case.",
"We present an algorithm for testing the k-vertex-connectivity of graphs with the given maximum degree. The time complexity of the algorithm is independent of the number of vertices and edges of graphs. Fixed degree bound d, a graph G with n vertices and a maximum degree at most d is called e-far from k-vertex-connectivity when at least @math edges must be added to or removed from G to obtain a k-vertex-connected graph with a maximum degree at most d. The algorithm always accepts every graph that is k-vertex-connected and rejects every graph that is e-far from k-vertex-connectivity with a probability of at least 2 3. The algorithm runs in @math time (c>1 is a constant) for (k−1)-vertex-connected graphs, and in @math time (c>1 is a constant) for general graphs. It is the first constant-time k-vertex-connectivity testing algorithm for general k≥4.",
"",
"",
"We present a probabilistic algorithm that, given a connected graph G (represented by adjacency lists) of average degree d, with edge weights in the set 1,...,w , and given a parameter @math , estimates in time @math the weight of the minimum spanning tree (MST) of G with a relative error of at most @math . Note that the running time does not depend on the number of vertices in G. We also prove a nearly matching lower bound of @math on the probe and time complexity of any approximation algorithm for MST weight. The essential component of our algorithm is a procedure for estimating in time @math the number of connected components of an unweighted graph to within an additive error of @math . (This becomes @math for @math .) The time bound is shown to be tight up to within the @math factor. Our connected-components algorithm picks @math vertices in the graph and then grows \"local spanning trees\" whose sizes are specified by a stochastic process. From the local information collected in this way, the algorithm is able to infer, with high confidence, an estimate of the number of connected components. We then show how estimates on the number of components in various subgraphs of G can be used to estimate the weight of its MST.",
"",
"For a given graph G over n vertices, let OPT\"G denote the size of an optimal solution in G of a particular minimization problem (e.g., the size of a minimum vertex cover). A randomized algorithm will be called an @a-approximation algorithm with an additive error for this minimization problem if for any given additive error parameter @e>0 it computes a value [email protected]? such that, with probability at least 2 3, it holds that OPT\"[email protected][email protected][email protected][email protected]@?OPT\"[email protected] Assume that the maximum degree or average degree of G is bounded. In this case, we show a reduction from local distributed approximation algorithms for the vertex cover problem to sublinear approximation algorithms for this problem. This reduction can be modified easily and applied to other optimization problems that have local distributed approximation algorithms, such as the dominating set problem. We also show that for the minimum vertex cover problem, the query complexity of such approximation algorithms must grow at least linearly with the average degree [email protected]? of the graph. This lower bound holds for every multiplicative factor @a and small constant @e as long as [email protected]?=O(n @a). In particular this means that for dense graphs it is not possible to design an algorithm whose complexity is o(n).",
"We further develop the study of testing graph properties as initiated by Goldreich, Goldwasser and Ron. Loosely speaking, given an oracle access to a graph, we wish to distinguish the case when the graph has a pre-determined property from the case when it is far from having this property. Whereas they view graphs as represented by their adjacency matrix and measure the distance between graphs as a fraction of all possible vertex pairs, we view graphs as represented by bounded-length incidence lists and measure the distance between graphs as a fraction of the maximum possible number of edges. Thus, while the previous model is most appropriate for the study of dense graphs, our model is most appropriate for the study of bounded-degree graphs. In particular, we present randomized algorithms for testing whether an unknown bounded-degree graph is connected, k-connected (for k > 1), cycle-free and Eulerian. Our algorithms work in time polynomial in 1 e, always accept the graph when it has the tested property, and reject with high probability if the graph is s-far from having the property. For example, the 2-connectivity algorithm rejects (with high probability) any N-vertex d-degree graph for which more than edN edges need to be added in order to make the graph 2-edge-connected. In addition we prove lower bounds of Ω(√N) on the query complexity of testing algorithms for the bipartite and expander properties."
]
} |
1707.07256 | 2739031953 | In this paper, we address the problem of person re-identification, which refers to associating the persons captured from different cameras. We propose a simple yet effective human part-aligned representation for handling the body part misalignment problem. Our approach decomposes the human body into regions (parts) which are discriminative for person matching, accordingly computes the representations over the regions, and aggregates the similarities computed between the corresponding regions of a pair of probe and gallery images as the overall matching score. Our formulation, inspired by attention models, is a deep neural network modeling the three steps together, which is learnt through minimizing the triplet loss function without requiring body part labeling information. Unlike most existing deep learning algorithms that learn a global or spatial partition-based local representation, our approach performs human body partition, and thus is more robust to pose changes and various human spatial distributions in the person bounding box. Our approach shows state-of-the-art results over standard datasets, Market- @math , CUHK @math , CUHK @math and VIPeR. | The network typically consists of two subnetworks: one for feature extraction and the other for matching. The feature extraction subnetwork could be simply ( ) a shallow network @cite_22 with one or two convolutional and max-pooling layers for feature extraction, or ( ) a deep network, e.g., VGGNet and its variants @cite_31 @cite_8 and GoogLeNet @cite_42 @cite_6 , which are pretrained over ImageNet and fine-tuned for person re-identification. The feature representation can be ( ) a global feature, e.g., the output of the fully-connected layer @cite_23 @cite_40 , which does not explicitly model the spatial information, or ( ) a combination (e.g., concatenation @cite_25 @cite_48 or contextual fusion @cite_75 ) of the features over regions, e.g., horizontal stripes @cite_25 @cite_48 @cite_75 , or grid cells @cite_22 @cite_64 , which are favorable for the later matching process to handle body part misalignment. Besides, the cross-dataset information @cite_39 is also exploited to learn an effective representation. | {
"cite_N": [
"@cite_64",
"@cite_22",
"@cite_8",
"@cite_48",
"@cite_42",
"@cite_6",
"@cite_39",
"@cite_40",
"@cite_23",
"@cite_31",
"@cite_75",
"@cite_25"
],
"mid": [
"",
"1982925187",
"",
"",
"2950179405",
"",
"2342611082",
"",
"2253171278",
"1686810756",
"",
""
],
"abstract": [
"",
"Person re-identification is to match pedestrian images from disjoint camera views detected by pedestrian detectors. Challenges are presented in the form of complex variations of lightings, poses, viewpoints, blurring effects, image resolutions, camera settings, occlusions and background clutter across camera views. In addition, misalignment introduced by the pedestrian detector will affect most existing person re-identification methods that use manually cropped pedestrian images and assume perfect detection. In this paper, we propose a novel filter pairing neural network (FPNN) to jointly handle misalignment, photometric and geometric transforms, occlusions and background clutter. All the key components are jointly optimized to maximize the strength of each component when cooperating with others. In contrast to existing works that use handcrafted features, our method automatically learns features optimal for the re-identification task from data. The learned filter pairs encode photometric transforms. Its deep architecture makes it possible to model a mixture of complex photometric and geometric transforms. We build the largest benchmark re-id dataset with 13, 164 images of 1, 360 pedestrians. Unlike existing datasets, which only provide manually cropped pedestrian images, our dataset provides automatically detected bounding boxes for evaluation close to practical applications. Our neural network significantly outperforms state-of-the-art methods on this dataset.",
"",
"",
"We propose a deep convolutional neural network architecture codenamed \"Inception\", which was responsible for setting the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC 2014). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. This was achieved by a carefully crafted design that allows for increasing the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC 2014 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.",
"",
"Learning generic and robust feature representations with data from multiple domains for the same problem is of great value, especially for the problems that have multiple datasets but none of them are large enough to provide abundant data variations. In this work, we present a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks (CNNs). When training a CNN with data from all the domains, some neurons learn representations shared across several domains, while some others are effective only for a specific one. Based on this important observation, we propose a Domain Guided Dropout algorithm to improve the feature learning procedure. Experiments show the effectiveness of our pipeline and the proposed algorithm. Our methods on the person re-identification problem outperform stateof-the-art methods on multiple datasets by large margins.",
"",
"This paper proposes a novel approach to person re-identification, a fundamental task in distributed multi-camera surveillance systems. Although a variety of powerful algorithms have been presented in the past few years, most of them usually focus on designing hand-crafted features and learning metrics either individually or sequentially. Different from previous works, we formulate a unified deep ranking framework that jointly tackles both of these key components to maximize their strengths. We start from the principle that the correct match of the probe image should be positioned in the top rank within the whole gallery set. An effective learning-to-rank algorithm is proposed to minimize the cost corresponding to the ranking disorders of the gallery. The ranking model is solved with a deep convolutional neural network (CNN) that builds the relation between input image pairs and their similarity scores through joint representation learning directly from raw image pixels. The proposed framework allows us to get rid of feature engineering and does not rely on any assumption. An extensive comparative evaluation is given, demonstrating that our approach significantly outperforms all the state-of-the-art approaches, including both traditional and CNN-based methods on the challenging VIPeR, CUHK-01, and CAVIAR4REID datasets. In addition, our approach has better ability to generalize across datasets without fine-tuning.",
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
"",
""
]
} |
1707.07256 | 2739031953 | In this paper, we address the problem of person re-identification, which refers to associating the persons captured from different cameras. We propose a simple yet effective human part-aligned representation for handling the body part misalignment problem. Our approach decomposes the human body into regions (parts) which are discriminative for person matching, accordingly computes the representations over the regions, and aggregates the similarities computed between the corresponding regions of a pair of probe and gallery images as the overall matching score. Our formulation, inspired by attention models, is a deep neural network modeling the three steps together, which is learnt through minimizing the triplet loss function without requiring body part labeling information. Unlike most existing deep learning algorithms that learn a global or spatial partition-based local representation, our approach performs human body partition, and thus is more robust to pose changes and various human spatial distributions in the person bounding box. Our approach shows state-of-the-art results over standard datasets, Market- @math , CUHK @math , CUHK @math and VIPeR. | The matching subnetwork can simply be a loss layer that penalizes the misalignment between learnt similarities and ground-truth similarities, e.g., pairwise loss @cite_25 @cite_72 @cite_66 @cite_64 @cite_74 , triplet loss and its variants @cite_4 @cite_20 @cite_61 @cite_41 . Besides using the off-the-shelf similarity function @cite_25 @cite_72 @cite_20 , e.g., cosine similarity or Euclidean distance, for comparing the feature representation, specific matching schemes are designed to eliminate the influence from body part misalignment. For instance, a matching subnetwork conducts convolution and max pooling operations, over the differences @cite_44 or the concatenation @cite_22 @cite_7 of the representations over grid cells of a pair of person images, to handle the misalignment problem. The approach with so called single-image and cross-image representations @cite_41 essentially combines the off-the-shelf distance and the matching network handling the misalignment. Instead of only matching the images over the final representation, the matching map in the intermediate features is used to guide the feature extraction in the later layers through a gated CNN @cite_19 . | {
"cite_N": [
"@cite_61",
"@cite_64",
"@cite_4",
"@cite_22",
"@cite_7",
"@cite_41",
"@cite_66",
"@cite_44",
"@cite_19",
"@cite_72",
"@cite_74",
"@cite_25",
"@cite_20"
],
"mid": [
"",
"",
"1971955426",
"1982925187",
"2578137651",
"",
"",
"1928419358",
"",
"",
"2519373641",
"",
""
],
"abstract": [
"",
"",
"Identifying the same individual across different scenes is an important yet difficult task in intelligent video surveillance. Its main difficulty lies in how to preserve similarity of the same person against large appearance and structure variation while discriminating different individuals. In this paper, we present a scalable distance driven feature learning framework based on the deep neural network for person re-identification, and demonstrate its effectiveness to handle the existing challenges. Specifically, given the training images with the class labels (person IDs), we first produce a large number of triplet units, each of which contains three images, i.e. one person with a matched reference and a mismatched reference. Treating the units as the input, we build the convolutional neural network to generate the layered representations, and follow with the L 2 distance metric. By means of parameter optimization, our framework tends to maximize the relative distance between the matched pair and the mismatched pair for each triplet unit. Moreover, a nontrivial issue arising with the framework is that the triplet organization cubically enlarges the number of training triplets, as one image can be involved into several triplet units. To overcome this problem, we develop an effective triplet generation scheme and an optimized gradient descent algorithm, making the computational load mainly depend on the number of original images instead of the number of triplets. On several challenging databases, our approach achieves very promising results and outperforms other state-of-the-art approaches. HighlightsWe present a novel feature learning framework for person re-identification.Our framework is based on the maximum relative distance comparison.The learning algorithm is scalable to process large amount of data.We demonstrate superior performances over other state-of-the-arts.",
"Person re-identification is to match pedestrian images from disjoint camera views detected by pedestrian detectors. Challenges are presented in the form of complex variations of lightings, poses, viewpoints, blurring effects, image resolutions, camera settings, occlusions and background clutter across camera views. In addition, misalignment introduced by the pedestrian detector will affect most existing person re-identification methods that use manually cropped pedestrian images and assume perfect detection. In this paper, we propose a novel filter pairing neural network (FPNN) to jointly handle misalignment, photometric and geometric transforms, occlusions and background clutter. All the key components are jointly optimized to maximize the strength of each component when cooperating with others. In contrast to existing works that use handcrafted features, our method automatically learns features optimal for the re-identification task from data. The learned filter pairs encode photometric transforms. Its deep architecture makes it possible to model a mixture of complex photometric and geometric transforms. We build the largest benchmark re-id dataset with 13, 164 images of 1, 360 pedestrians. Unlike existing datasets, which only provide manually cropped pedestrian images, our dataset provides automatically detected bounding boxes for evaluation close to practical applications. Our neural network significantly outperforms state-of-the-art methods on this dataset.",
"In this paper, we propose an end-to-end deep correspondence structure learning (DCSL) approach to address the cross-camera person-matching problem in the person re-identification task. The proposed DCSL approach captures the intrinsic structural information on persons by learning a semantics-aware image representation based on convolutional neural networks, which adaptively learns discriminative features for person identification. Furthermore, the proposed DCSL approach seeks to adaptively learn a hierarchical data-driven feature matching function which outputs the matching correspondence results between the learned semantics-aware image representations for a person pair. Finally, we set up a unified end-to-end deep learning scheme to jointly optimize the processes of semantics-aware image representation learning and cross-person correspondence structure learning, leading to more reliable and robust person re-identification results in complicated scenarios. Experimental results on several benchmark datasets demonstrate the effectiveness of our approach against the state-of-the-art approaches.",
"",
"",
"In this work, we propose a method for simultaneously learning features and a corresponding similarity metric for person re-identification. We present a deep convolutional architecture with layers specially designed to address the problem of re-identification. Given a pair of images as input, our network outputs a similarity value indicating whether the two input images depict the same person. Novel elements of our architecture include a layer that computes cross-input neighborhood differences, which capture local relationships between the two input images based on mid-level features from each input image. A high-level summary of the outputs of this layer is computed by a layer of patch summary features, which are then spatially integrated in subsequent layers. Our method significantly outperforms the state of the art on both a large data set (CUHK03) and a medium-sized data set (CUHK01), and is resistant to over-fitting. We also demonstrate that by initially training on an unrelated large data set before fine-tuning on a small target data set, our network can achieve results comparable to the state of the art even on a small data set (VIPeR).",
"",
"",
"Person re-identification is challenging due to the large variations of pose, illumination, occlusion and camera view. Owing to these variations, the pedestrian data is distributed as highly-curved manifolds in the feature space, despite the current convolutional neural networks (CNN)’s capability of feature extraction. However, the distribution is unknown, so it is difficult to use the geodesic distance when comparing two samples. In practice, the current deep embedding methods use the Euclidean distance for the training and test. On the other hand, the manifold learning methods suggest to use the Euclidean distance in the local range, combining with the graphical relationship between samples, for approximating the geodesic distance. From this point of view, selecting suitable positive (i.e. intra-class) training samples within a local range is critical for training the CNN embedding, especially when the data has large intra-class variations. In this paper, we propose a novel moderate positive sample mining method to train robust CNN for person re-identification, dealing with the problem of large variation. In addition, we improve the learning by a metric weight constraint, so that the learned metric has a better generalization ability. Experiments show that these two strategies are effective in learning robust deep metrics for person re-identification, and accordingly our deep model significantly outperforms the state-of-the-art methods on several benchmarks of person re-identification. Therefore, the study presented in this paper may be useful in inspiring new designs of deep models for person re-identification.",
"",
""
]
} |
1707.07256 | 2739031953 | In this paper, we address the problem of person re-identification, which refers to associating the persons captured from different cameras. We propose a simple yet effective human part-aligned representation for handling the body part misalignment problem. Our approach decomposes the human body into regions (parts) which are discriminative for person matching, accordingly computes the representations over the regions, and aggregates the similarities computed between the corresponding regions of a pair of probe and gallery images as the overall matching score. Our formulation, inspired by attention models, is a deep neural network modeling the three steps together, which is learnt through minimizing the triplet loss function without requiring body part labeling information. Unlike most existing deep learning algorithms that learn a global or spatial partition-based local representation, our approach performs human body partition, and thus is more robust to pose changes and various human spatial distributions in the person bounding box. Our approach shows state-of-the-art results over standard datasets, Market- @math , CUHK @math , CUHK @math and VIPeR. | Our human body part estimation scheme is inspired by the attention model that is successfully applied to many applications such as image captioning @cite_3 . Compared to the work @cite_29 that is based on attention models and LSTM, our approach is simple and easily implemented, and empirical results show that our approach performs better. | {
"cite_N": [
"@cite_29",
"@cite_3"
],
"mid": [
"2432402544",
"2950178297"
],
"abstract": [
"Person re-identification across disjoint camera views has been widely applied in video surveillance yet it is still a challenging problem. One of the major challenges lies in the lack of spatial and temporal cues, which makes it difficult to deal with large variations of lighting conditions, viewing angles, body poses, and occlusions. Recently, several deep-learning-based person re-identification approaches have been proposed and achieved remarkable performance. However, most of those approaches extract discriminative features from the whole frame at one glimpse without differentiating various parts of the persons to identify. It is essentially important to examine multiple highly discriminative local regions of the person images in details through multiple glimpses for dealing with the large appearance variance. In this paper, we propose a new soft attention-based model, i.e. , the end-to-end comparative attention network (CAN), specifically tailored for the task of person re-identification. The end-to-end CAN learns to selectively focus on parts of pairs of person images after taking a few glimpses of them and adaptively comparing their appearance. The CAN model is able to learn which parts of images are relevant for discerning persons and automatically integrates information from different parts to determine whether a pair of images belongs to the same person. In other words, our proposed CAN model simulates the human perception process to verify whether two images are from the same person. Extensive experiments on four benchmark person re-identification data sets, including CUHK01, CHUHK03, Market-1501, and VIPeR, clearly demonstrate that our proposed end-to-end CAN for person re-identification outperforms well established baselines significantly and offer the new state-of-the-art performance.",
"Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr8k, Flickr30k and MS COCO."
]
} |
1707.07250 | 2738581557 | Multimodal sentiment analysis is an increasingly popular research area, which extends the conventional language-based definition of sentiment analysis to a multimodal setup where other relevant modalities accompany language. In this paper, we pose the problem of multimodal sentiment analysis as modeling intra-modality and inter-modality dynamics. We introduce a novel model, termed Tensor Fusion Network, which learns both such dynamics end-to-end. The proposed approach is tailored for the volatile nature of spoken language in online videos as well as accompanying gestures and voice. In the experiments, our model outperforms state-of-the-art approaches for both multimodal and unimodal sentiment analysis. | is an emerging research area that integrates verbal and nonverbal behaviors into the detection of user sentiment. There exist several multimodal datasets that include sentiment annotations, including the newly-introduced CMU-MOSI dataset @cite_7 , as well as other datasets including ICT-MMMO @cite_55 , YouTube @cite_12 , and MOUD @cite_53 , however CMU-MOSI is the only English dataset with utterance-level sentiment labels. The newest multimodal sentiment analysis approaches have used deep neural networks, including convolutional neural networks (CNNs) with multiple-kernel learning @cite_25 , SAL-CNN @cite_33 which learns generalizable features across speakers, and support vector machines (SVMs) with a multimodal dictionary @cite_14 . | {
"cite_N": [
"@cite_14",
"@cite_33",
"@cite_7",
"@cite_55",
"@cite_53",
"@cite_25",
"@cite_12"
],
"mid": [
"2244706744",
"2522893818",
"2556418146",
"",
"2098689807",
"2251394420",
"2122563357"
],
"abstract": [
"There has been substantial progress in the field of text based sentiment analysis but little effort has been made to incorporate other modalities. Previous work in sentiment analysis has shown that using multimodal data yields to more accurate models of sentiment. Efforts have been made towards expressing sentiment as a spectrum of intensity rather than just positive or negative. Such models are useful not only for detection of positivity or negativity, but also giving out a score of how positive or negative a statement is. Based on the state of the art studies in sentiment analysis, prediction in terms of sentiment score is still far from accurate, even in large datasets [27]. Another challenge in sentiment analysis is dealing with small segments or micro opinions as they carry less context than large segments thus making analysis of the sentiment harder. This paper presents a Ph.D. thesis shaped towards comprehensive studies in multimodal micro-opinion sentiment intensity analysis.",
"Multimodal sentiment analysis is drawing an increasing amount of attention these days. It enables mining of opinions in video reviews and surveys which are now available aplenty on online platforms like YouTube. However, the limited number of high-quality multimodal sentiment data samples may introduce the problem of the sentiment being dependent on the individual specific features in the dataset. This results in a lack of generalizability of the trained models for classification on larger online platforms. In this paper, we first examine the data and verify the existence of this dependence problem. Then we propose a Select-Additive Learning (SAL) procedure that improves the generalizability of trained discriminative neural networks. SAL is a two-phase learning method. In Selection phase, it selects the confounding learned representation. In Addition phase, it forces the classifier to discard confounded representations by adding Gaussian noise. In our experiments, we show how SAL improves the generalizability of state-of-the-art models. We increase prediction accuracy significantly in all three modalities (text, audio, video), as well as in their fusion. We show how SAL, even when trained on one dataset, achieves good accuracy across test datasets.",
"People share their opinions, stories, and reviews through online video sharing websites every day. The automatic analysis of these online opinion videos is bringing new or understudied research challenges to the field of computational linguistics and multimodal analysis. Among these challenges is the fundamental question of exploiting the dynamics between visual gestures and verbal messages to be able to better model sentiment. This article addresses this question in four ways: introducing the first multimodal dataset with opinion-level sentiment intensity annotations; studying the prototypical interaction patterns between facial gestures and spoken words when inferring sentiment intensity; proposing a new computational representation, called multimodal dictionary, based on a language-gesture study; and evaluating the authors' proposed approach in a speaker-independent paradigm for sentiment intensity prediction. The authors' study identifies four interaction types between facial gestures and verbal content: neutral, emphasizer, positive, and negative interactions. Experiments show statistically significant improvement when using multimodal dictionary representation over the conventional early fusion representation (that is, feature concatenation).",
"",
"During real-life interactions, people are naturally gesturing and modulating their voice to emphasize specific points or to express their emotions. With the recent growth of social websites such as YouTube, Facebook, and Amazon, video reviews are emerging as a new source of multimodal and natural opinions that has been left almost untapped by automatic opinion analysis techniques. This paper presents a method for multimodal sentiment classification, which can identify the sentiment expressed in utterance-level visual datastreams. Using a new multimodal dataset consisting of sentiment annotated utterances extracted from video reviews, we show that multimodal sentiment analysis can be effectively performed, and that the joint use of visual, acoustic, and linguistic modalities can lead to error rate reductions of up to 10.5 as compared to the best performing individual modality.",
"We present a novel way of extracting features from short texts, based on the activation values of an inner layer of a deep convolutional neural network. We use the extracted features in multimodal sentiment analysis of short video clips representing one sentence each. We use the combined feature vectors of textual, visual, and audio modalities to train a classifier based on multiple kernel learning, which is known to be good at heterogeneous data. We obtain 14 performance improvement over the state of the art and present a parallelizable decision-level data fusion method, which is much faster, though slightly less accurate.",
"With more than 10,000 new videos posted online every day on social websites such as YouTube and Facebook, the internet is becoming an almost infinite source of information. One crucial challenge for the coming decade is to be able to harvest relevant information from this constant flow of multimodal data. This paper addresses the task of multimodal sentiment analysis, and conducts proof-of-concept experiments that demonstrate that a joint model that integrates visual, audio, and textual features can be effectively used to identify sentiment in Web videos. This paper makes three important contributions. First, it addresses for the first time the task of tri-modal sentiment analysis, and shows that it is a feasible task that can benefit from the joint exploitation of visual, audio and textual modalities. Second, it identifies a subset of audio-visual features relevant to sentiment analysis and present guidelines on how to integrate these features. Finally, it introduces a new dataset consisting of real online data, which will be useful for future research in this area."
]
} |
1707.07410 | 2738401084 | We present a point tracking system powered by two deep convolutional neural networks. The first network, MagicPoint, operates on single images and extracts salient 2D points. The extracted points are "SLAM-ready" because they are by design isolated and well-distributed throughout the image. We compare this network against classical point detectors and discover a significant performance gap in the presence of image noise. As transformation estimation is more simple when the detected points are geometrically stable, we designed a second network, MagicWarp, which operates on pairs of point images (outputs of MagicPoint), and estimates the homography that relates the inputs. This transformation engine differs from traditional approaches because it does not use local point descriptors, only point locations. Both networks are trained with simple synthetic data, alleviating the requirement of expensive external camera ground truthing and advanced graphics rendering pipelines. The system is fast and lean, easily running 30+ FPS on a single CPU. | Individual components of SLAM systems have recently been tackled with supervised deep learning methods. The feature point detection and description stage was tackled in the work of @cite_14 , where a convolutional neural network was trained using image patches filtered through a classical Structure From Motion pipeline. Transformation estimation was shown to be done successfully by CNNs in @cite_16 , where a deep network was trained on a large dataset of warped natural images. The transformation estimation done in this work was direct, meaning that the convolutional neural network directly mapped pairs of images to their transforms. The work of @cite_4 tackled dense optical flow. The problem of camera localization was also tackled with a CNN in @cite_2 , where a network was trained to learn a mapping from images to absolute 6DOF poses. A deep version of the RANSAC algorithm was presented in @cite_20 , where a deep network was trained to learn a robust estimator for camera localization. There have also been many works such as @cite_0 in the direction of deep relocalization via metric learning, where two images with similar pose are mapped to a similar point on an embedding produced by a deep network. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_0",
"@cite_2",
"@cite_16",
"@cite_20"
],
"mid": [
"",
"",
"2262703527",
"2951336016",
"2439114332",
"2950681848"
],
"abstract": [
"",
"",
"Place recognition is one of the most challenging problems in computer vision, and has become a key part in mobile robotics and autonomous driving applications for performing loop closure in visual SLAM systems. Moreover, the difficulty of recognizing a revisited location increases with appearance changes caused, for instance, by weather or illumination variations, which hinders the long-term application of such algorithms in real environments. In this paper we present a convolutional neural network (CNN), trained for the first time with the purpose of recognizing revisited locations under severe appearance changes, which maps images to a low dimensional space where Euclidean distances represent place dissimilarity. In order for the network to learn the desired invariances, we train it with triplets of images selected from datasets which present a challenging variability in visual appearance. The triplets are selected in such way that two samples are from the same location and the third one is taken from a different place. We validate our system through extensive experimentation, where we demonstrate better performance than state-of-art algorithms in a number of popular datasets.",
"We present a robust and real-time monocular six degree of freedom relocalization system. Our system trains a convolutional neural network to regress the 6-DOF camera pose from a single RGB image in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking 5ms per frame to compute. It obtains approximately 2m and 6 degree accuracy for large scale outdoor scenes and 0.5m and 10 degree accuracy indoors. This is achieved using an efficient 23 layer deep convnet, demonstrating that convnets can be used to solve complicated out of image plane regression problems. This was made possible by leveraging transfer learning from large scale classification data. We show the convnet localizes from high level features and is robust to difficult lighting, motion blur and different camera intrinsics where point based SIFT registration fails. Furthermore we show how the pose feature that is produced generalizes to other scenes allowing us to regress pose with only a few dozen training examples. PoseNet code, dataset and an online demonstration is available on our project webpage, at this http URL",
"We present a deep convolutional neural network for estimating the relative homography between a pair of images. Our feed-forward network has 10 layers, takes two stacked grayscale images as input, and produces an 8 degree of freedom homography which can be used to map the pixels from the first image to the second. We present two convolutional neural network architectures for HomographyNet: a regression network which directly estimates the real-valued homography parameters, and a classification network which produces a distribution over quantized homographies. We use a 4-point homography parameterization which maps the four corners from one image into the second image. Our networks are trained in an end-to-end fashion using warped MS-COCO images. Our approach works without the need for separate local feature detection and transformation estimation stages. Our deep models are compared to a traditional homography estimator based on ORB features and we highlight the scenarios where HomographyNet outperforms the traditional technique. We also describe a variety of applications powered by deep homography estimation, thus showcasing the flexibility of a deep learning approach.",
"RANSAC is an important algorithm in robust optimization and a central building block for many computer vision applications. In recent years, traditionally hand-crafted pipelines have been replaced by deep learning pipelines, which can be trained in an end-to-end fashion. However, RANSAC has so far not been used as part of such deep learning pipelines, because its hypothesis selection procedure is non-differentiable. In this work, we present two different ways to overcome this limitation. The most promising approach is inspired by reinforcement learning, namely to replace the deterministic hypothesis selection by a probabilistic selection for which we can derive the expected loss w.r.t. to all learnable parameters. We call this approach DSAC, the differentiable counterpart of RANSAC. We apply DSAC to the problem of camera localization, where deep learning has so far failed to improve on traditional approaches. We demonstrate that by directly minimizing the expected loss of the output camera poses, robustly estimated by RANSAC, we achieve an increase in accuracy. In the future, any deep learning pipeline can use DSAC as a robust optimization component."
]
} |
1707.07410 | 2738401084 | We present a point tracking system powered by two deep convolutional neural networks. The first network, MagicPoint, operates on single images and extracts salient 2D points. The extracted points are "SLAM-ready" because they are by design isolated and well-distributed throughout the image. We compare this network against classical point detectors and discover a significant performance gap in the presence of image noise. As transformation estimation is more simple when the detected points are geometrically stable, we designed a second network, MagicWarp, which operates on pairs of point images (outputs of MagicPoint), and estimates the homography that relates the inputs. This transformation engine differs from traditional approaches because it does not use local point descriptors, only point locations. Both networks are trained with simple synthetic data, alleviating the requirement of expensive external camera ground truthing and advanced graphics rendering pipelines. The system is fast and lean, easily running 30+ FPS on a single CPU. | A series of works have also tackled multiple SLAM problems concurrently. In @cite_8 , a CNN was trained to simultaneous estimate the depth and motion of a monocular camera pair. Interestingly, the models did not work well when the network was trained on the two tasks separately, but worked much better when trained jointly. This observation is consistent with natural regularization of multi task learning. This approach relies on supervised training data for both motion and depth cues. | {
"cite_N": [
"@cite_8"
],
"mid": [
"2561074213"
],
"abstract": [
"In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training."
]
} |
1707.07410 | 2738401084 | We present a point tracking system powered by two deep convolutional neural networks. The first network, MagicPoint, operates on single images and extracts salient 2D points. The extracted points are "SLAM-ready" because they are by design isolated and well-distributed throughout the image. We compare this network against classical point detectors and discover a significant performance gap in the presence of image noise. As transformation estimation is more simple when the detected points are geometrically stable, we designed a second network, MagicWarp, which operates on pairs of point images (outputs of MagicPoint), and estimates the homography that relates the inputs. This transformation engine differs from traditional approaches because it does not use local point descriptors, only point locations. Both networks are trained with simple synthetic data, alleviating the requirement of expensive external camera ground truthing and advanced graphics rendering pipelines. The system is fast and lean, easily running 30+ FPS on a single CPU. | There is also a new direction of research at the intersection of deep learning and SLAM where no or very few ground truth measurements are required. By formulating loss functions to maximize photometric consistency, the works of @cite_1 and @cite_10 showed an ego-motion and depth estimation system based on image prediction. Work in this direction show that there is much promise in moving away from strong supervision. | {
"cite_N": [
"@cite_10",
"@cite_1"
],
"mid": [
"2520707372",
"2609883120"
],
"abstract": [
"Learning based methods have shown very promising results for the task of depth estimation in single images. However, most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training. Just recording quality depth data in a range of environments is a challenging problem. In this paper, we innovate beyond existing approaches, replacing the use of explicit depth data during training with easier-to-obtain binocular stereo footage. We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data. Ex-ploiting epipolar geometry constraints, we generate disparity images by training our network with an image reconstruction loss. We show that solving for image reconstruction alone results in poor quality depth images. To overcome this problem, we propose a novel training loss that enforces consistency between the disparities produced relative to both the left and right images, leading to improved performance and robustness compared to existing approaches. Our method produces state of the art results for monocular depth estimation on the KITTI driving dataset, even outperforming supervised methods that have been trained with ground truth depth.",
"We present an unsupervised learning framework for the task of monocular depth and camera motion estimation from unstructured video sequences. In common with recent work [10, 14, 16], we use an end-to-end learning approach with view synthesis as the supervisory signal. In contrast to the previous work, our method is completely unsupervised, requiring only monocular video sequences for training. Our method uses single-view depth and multiview pose networks, with a loss based on warping nearby views to the target using the computed depth and pose. The networks are thus coupled by the loss during training, but can be applied independently at test time. Empirical evaluation on the KITTI dataset demonstrates the effectiveness of our approach: 1) monocular depth performs comparably with supervised methods that use either ground-truth pose or depth for training, and 2) pose estimation performs favorably compared to established SLAM systems under comparable input settings."
]
} |
1707.07013 | 2737498590 | State-of-the-art Deep Neural Networks can be easily fooled into providing incorrect high-confidence predictions for images with small amounts of adversarial noise. Does this expose a flaw with deep neural networks, or do we simply need a better way to estimate confidence? In this paper we consider the problem of accurately estimating predictive confidence. We formulate this problem as that of density modelling, and show how traditional methods such as softmax produce poor estimates. To address this issue, we propose a novel confidence measure based on density modelling approaches. We test these measures on images distorted by blur, JPEG compression, random noise and adversarial noise. Experiments show that our confidence measure consistently shows reduced confidence scores in the presence of such distortions - a property which softmax often lacks. | Uncertainty or Confidence estimation has gained a lot of attention in recent times. Gal al @cite_15 presented a method of estimating the uncertainty in neural network model by performing dropout averaging of predictions during test time. Bendale al @cite_16 presented Open set deep networks, which attempt to determine whether a given image belongs to any of the classes it was trained for. However, both these methods use the softmax function to compute uncertainty. We will show shortly that uncertainty estimates from softmax contain certain pathologies, making them unsuitable for this task. | {
"cite_N": [
"@cite_15",
"@cite_16"
],
"mid": [
"582134693",
"2272331516"
],
"abstract": [
"Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty. In comparison, Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with a prohibitive computational cost. In this paper we develop a new theoretical framework casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes. A direct result of this theory gives us tools to model uncertainty with dropout NNs -- extracting information from existing models that has been thrown away so far. This mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy. We perform an extensive study of the properties of dropout's uncertainty. Various network architectures and non-linearities are assessed on tasks of regression and classification, using MNIST as an example. We show a considerable improvement in predictive log-likelihood and RMSE compared to existing state-of-the-art methods, and finish by using dropout's uncertainty in deep reinforcement learning.",
"Deep networks have produced significant gains for various visual recognition problems, leading to high impact academic and commercial applications. Recent work in deep networks highlighted that it is easy to generate images that humans would never classify as a particular object class, yet networks classify such images high confidence as that given class - deep network are easily fooled with images humans do not consider meaningful. The closed set nature of deep networks forces them to choose from one of the known classes leading to such artifacts. Recognition in the real world is open set, i.e. the recognition system should reject unknown unseen classes at test time. We present a methodology to adapt deep networks for open set recognition, by introducing a new model layer, OpenMax, which estimates the probability of an input being from an unknown class. A key element of estimating the unknown probability is adapting Meta-Recognition concepts to the activation patterns in the penultimate layer of the network. OpenMax allows rejection of \"fooling\" and unrelated open set images presented to the system; OpenMax greatly reduces the number of obvious errors made by a deep network. We prove that the OpenMax concept provides bounded open space risk, thereby formally providing an open set recognition solution. We evaluate the resulting open set deep networks using pre-trained networks from the Caffe Model-zoo on ImageNet 2012 validation data, and thousands of fooling and open set images. The proposed OpenMax model significantly outperforms open set recognition accuracy of basic deep networks as well as deep networks with thresholding of SoftMax probabilities."
]
} |
1707.07013 | 2737498590 | State-of-the-art Deep Neural Networks can be easily fooled into providing incorrect high-confidence predictions for images with small amounts of adversarial noise. Does this expose a flaw with deep neural networks, or do we simply need a better way to estimate confidence? In this paper we consider the problem of accurately estimating predictive confidence. We formulate this problem as that of density modelling, and show how traditional methods such as softmax produce poor estimates. To address this issue, we propose a novel confidence measure based on density modelling approaches. We test these measures on images distorted by blur, JPEG compression, random noise and adversarial noise. Experiments show that our confidence measure consistently shows reduced confidence scores in the presence of such distortions - a property which softmax often lacks. | Modern neural network architectures are sensitive to adversarial examples @cite_11 . These are images produced by the addition of small perturbations to correctly classified samples. Generating adversarial examples to the classifier is one of the active areas of research in the Deep Learning Community. Goodfellow al @cite_12 presented a method to generate adversarial examples and also showed that retraining the network with these examples can be used for regularization. Nyugen al @cite_5 also strengthened the claim that networks can be fooled easily by generating . Moosavi al @cite_13 presented an effective and fast way of generating the perturbations required to misclassify any image. Moosavi al @cite_14 also show that there exists universal adversarial perturbations given a classifier, that can be applied to any image for misclassification . Researchers have also shown that results of face recognition algorithms can be tampered by wearing a specific design of eyeglass frames @cite_4 . Such examples present a challenge to the idea of using neural networks for commercial applications, since security could be easily compromised. These can be overcome (in principle) by using a good uncertainty estimate of predictions. | {
"cite_N": [
"@cite_14",
"@cite_4",
"@cite_5",
"@cite_13",
"@cite_12",
"@cite_11"
],
"mid": [
"2953047670",
"2535873859",
"",
"2950159395",
"1945616565",
"1673923490"
],
"abstract": [
"Given a state-of-the-art deep neural network classifier, we show the existence of a universal (image-agnostic) and very small perturbation vector that causes natural images to be misclassified with high probability. We propose a systematic algorithm for computing universal perturbations, and show that state-of-the-art deep neural networks are highly vulnerable to such perturbations, albeit being quasi-imperceptible to the human eye. We further empirically analyze these universal perturbations and show, in particular, that they generalize very well across neural networks. The surprising existence of universal perturbations reveals important geometric correlations among the high-dimensional decision boundary of classifiers. It further outlines potential security breaches with the existence of single directions in the input space that adversaries can possibly exploit to break a classifier on most natural images.",
"Machine learning is enabling a myriad innovations, including new algorithms for cancer diagnosis and self-driving cars. The broad use of machine learning makes it important to understand the extent to which machine-learning algorithms are subject to attack, particularly when used in applications where physical security or safety is at risk. In this paper, we focus on facial biometric systems, which are widely used in surveillance and access control. We define and investigate a novel class of attacks: attacks that are physically realizable and inconspicuous, and allow an attacker to evade recognition or impersonate another individual. We develop a systematic method to automatically generate such attacks, which are realized through printing a pair of eyeglass frames. When worn by the attacker whose image is supplied to a state-of-the-art face-recognition algorithm, the eyeglasses allow her to evade being recognized or to impersonate another individual. Our investigation focuses on white-box face-recognition systems, but we also demonstrate how similar techniques can be used in black-box scenarios, as well as to avoid face detection.",
"",
"State-of-the-art deep neural networks have achieved impressive results on many image classification tasks. However, these same architectures have been shown to be unstable to small, well sought, perturbations of the images. Despite the importance of this phenomenon, no effective methods have been proposed to accurately compute the robustness of state-of-the-art deep classifiers to such perturbations on large-scale datasets. In this paper, we fill this gap and propose the DeepFool algorithm to efficiently compute perturbations that fool deep networks, and thus reliably quantify the robustness of these classifiers. Extensive experimental results show that our approach outperforms recent methods in the task of computing adversarial perturbations and making classifiers more robust.",
"Several machine learning models, including neural networks, consistently misclassify adversarial examples---inputs formed by applying small but intentionally worst-case perturbations to examples from the dataset, such that the perturbed input results in the model outputting an incorrect answer with high confidence. Early attempts at explaining this phenomenon focused on nonlinearity and overfitting. We argue instead that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature. This explanation is supported by new quantitative results while giving the first explanation of the most intriguing fact about them: their generalization across architectures and training sets. Moreover, this view yields a simple and fast method of generating adversarial examples. Using this approach to provide examples for adversarial training, we reduce the test set error of a maxout network on the MNIST dataset.",
"Deep neural networks are highly expressive models that have recently achieved state of the art performance on speech and visual recognition tasks. While their expressiveness is the reason they succeed, it also causes them to learn uninterpretable solutions that could have counter-intuitive properties. In this paper we report two such properties. First, we find that there is no distinction between individual high level units and random linear combinations of high level units, according to various methods of unit analysis. It suggests that it is the space, rather than the individual units, that contains of the semantic information in the high layers of neural networks. Second, we find that deep neural networks learn input-output mappings that are fairly discontinuous to a significant extend. We can cause the network to misclassify an image by applying a certain imperceptible perturbation, which is found by maximizing the network's prediction error. In addition, the specific nature of these perturbations is not a random artifact of learning: the same perturbation can cause a different network, that was trained on a different subset of the dataset, to misclassify the same input."
]
} |
1707.07075 | 2738026219 | The production of sports highlight packages summarizing a game's most exciting moments is an essential task for broadcast media. Yet, it requires labor-intensive video editing. We propose a novel approach for auto-curating sports highlights, and use it to create a real-world system for the editorial aid of golf highlight reels. Our method fuses information from the players' reactions (action recognition such as high-fives and fist pumps), spectators (crowd cheering), and commentator (tone of the voice and word analysis) to determine the most interesting moments of a game. We accurately identify the start and end frames of key shot highlights with additional metadata, such as the player's name and the hole number, allowing personalized content summarization and retrieval. In addition, we introduce new techniques for learning our classifiers with reduced manual training data annotation by exploiting the correlation of different modalities. Our work has been demonstrated at a major golf tournament, successfully extracting highlights from live video streams over four consecutive days. | Sports Highlights Generation. Several methods have been proposed to automatically extract highlights from sports videos based on audio and visual cues. Example approaches include the analysis of replays @cite_15 , crowd cheering @cite_16 , motion features @cite_2 , and closed captioning @cite_3 . More recently, @cite_26 used contextual cues from the environment to understand the excitement levels within a basketball game. Tang and Boring @cite_6 proposed to automatically produce highlights by analyzing social media services such as twitter. @cite_18 developed a method for forecasting sports highlights to achieve more effective coverage of multiple games happening at the same time. Different from existing methods, our proposed approach offers a unique combination of excitement measures to produce highlights, including information from the spectators , the commentator , and the player reaction. In addition, we enable personalized highlight generation or retrieval based on a viewer s favorite players. | {
"cite_N": [
"@cite_18",
"@cite_26",
"@cite_3",
"@cite_6",
"@cite_2",
"@cite_15",
"@cite_16"
],
"mid": [
"2582896079",
"2949078144",
"2029545460",
"2117849405",
"2128889560",
"2120748909",
"2142524229"
],
"abstract": [
"",
"The massive growth of sports videos has resulted in a need for automatic generation of sports highlights that are comparable in quality to the hand-edited highlights produced by broadcasters such as ESPN. Unlike previous works that mostly use audio-visual cues derived from the video, we propose an approach that additionally leverages contextual cues derived from the environment that the game is being played in. The contextual cues provide information about the excitement levels in the game, which can be ranked and selected to automatically produce high-quality basketball highlights. We introduce a new dataset of 25 NCAA games along with their play-by-play stats and the ground-truth excitement data for each basket. We explore the informativeness of five different cues derived from the video and from the environment through user studies. Our experiments show that for our study participants, the highlights produced by our system are comparable to the ones produced by ESPN for the same games.",
"We have developed a novel system for baseball video event detection and summarization using superimposed caption text detection and recognition. The system detects different types of semantic level events in baseball video including scoring and last pitch of each batter. The system has two components: event detection and event boundary detection. Event detection is realized by change detection and recognition of game stat texts (such as text information showing in score box). Event boundary detection is achieved using our previously developed algorithm, which detects the pitch view as the event beginning and nonactive view as potential endings of the event. One unique contribution of the system is its capability to accurately detect the semantic level events by combining video text recognition with camera view recognition. Another unique feature is the real-time processing speed by taking advantage of compressed-domain approaches in part of the algorithms such as caption detection. To the best of our knowledge, this is the first system achieving accurate detection of multiple types of high-level semantic events in baseball videos.",
"During a live sports event, many sports fans use social media as a part of their viewing experience, reporting on their thoughts on the event as it unfolds. In this work, we use this information stream to semantically annotate live broadcast sports games, using these annotations to select video highlights from the game. We demonstrate that this approach can be used to select highlights specific for fans of each team, and that these clips reflect the emotions of a fan during a game. Further, we describe how these clips differ from those seen on nightly sportscasts.",
"In our past work we have used temporal patterns of motion activity to extract sports highlights. We have also used audio classification based approaches to develop a common audio-based platform for feature extraction that works across three different sports. In this paper, we combine the two aforementioned complementary approaches so as to get higher accuracy. We propose a framework for mining the semantic audio-visual labels in order to detect \"interesting\" events. Our results show that the proposed techniques work well across our three sports of interest, soccer, golf and baseball.",
"Highlight summarization technology has been studied widely in sports video analysis. In this paper, we propose a highlight summarization system based on replays. First the replay clips in the sports video are extracted as the highlight candidates. Then the features including audio energy and motion activity are employed to rank the arousal level of the replay clips. Finally we model the highlight with the arousal rank to generate summarization. The contribution of this paper concentrates on two aspects. Firstly, Event-Replay (ER) structure is proposed and some new features are employed to represent the arousal levels of ER for general sports video. Secondly a novel highlight model is proposed considering the inter-relation of ERs. The experiments evaluate the rationality of the system.",
"We have developed a unified framework to extract highlights from three sports - baseball, golf and soccer - by detecting some of the common audio events that are directly indicative of highlights. We used MPEG-7 audio features and entropic prior hidden Markov models (HMM) for feature extraction and classification, respectively, to recognize these common audio events. Together with pre- and post-processing techniques using general sports knowledge, we have been able to generate promising results dealing with an audio track that is dominated by audio mixtures and noisy background."
]
} |
1707.07075 | 2738026219 | The production of sports highlight packages summarizing a game's most exciting moments is an essential task for broadcast media. Yet, it requires labor-intensive video editing. We propose a novel approach for auto-curating sports highlights, and use it to create a real-world system for the editorial aid of golf highlight reels. Our method fuses information from the players' reactions (action recognition such as high-fives and fist pumps), spectators (crowd cheering), and commentator (tone of the voice and word analysis) to determine the most interesting moments of a game. We accurately identify the start and end frames of key shot highlights with additional metadata, such as the player's name and the hole number, allowing personalized content summarization and retrieval. In addition, we introduce new techniques for learning our classifiers with reduced manual training data annotation by exploiting the correlation of different modalities. Our work has been demonstrated at a major golf tournament, successfully extracting highlights from live video streams over four consecutive days. | Self-Supervised Learning. In recent years, there has been significant interest in methods that learn deep neural network classifiers without requiring a large amount of manually annotated training examples. In particular, self-supervised learning approaches rely on auxiliary tasks for feature learning, leveraging sources of supervision that are usually available for free'' and in large quantities to regularize deep neural network models. Examples of auxiliary tasks include the prediction of ego-motion @cite_11 @cite_0 , location and weather @cite_14 , spatial context or patch layout @cite_8 @cite_24 , image colorization @cite_12 , and temporal coherency @cite_19 . @cite_9 explored the natural synchronization between vision and sound to learn an acoustic representation from unlabeled video. We leverage this work to build audio models for crowd cheering and commentator excitement with a few training examples, and use these classifiers to constrain the training data collection for player reaction recognition. More interestingly, we exploit the detection of TV graphics as a free supervisory signal to learn feature representations for player recognition from unlabeled video. | {
"cite_N": [
"@cite_14",
"@cite_8",
"@cite_9",
"@cite_0",
"@cite_24",
"@cite_19",
"@cite_12",
"@cite_11"
],
"mid": [
"2337242548",
"2321533354",
"2544224704",
"2198618282",
"2342877626",
"2145038566",
"2326925005",
"2951590555"
],
"abstract": [
"The way people look in terms of facial attributes (ethnicity, hair color, facial hair, etc.) and the clothes or accessories they wear (sunglasses, hat, hoodies, etc.) is highly dependent on geo-location and weather condition, respectively. This work explores, for the first time, the use of this contextual information, as people with wearable cameras walk across different neighborhoods of a city, in order to learn a rich feature representation for facial attribute classification, without the costly manual annotation required by previous methods. By tracking the faces of casual walkers on more than 40 hours of egocentric video, we are able to cover tens of thousands of different identities and automatically extract nearly 5 million pairs of images connected by or from different face tracks, along with their weather and location context, under pose and lighting variations. These image pairs are then fed into a deep network that preserves similarity of images connected by the same track, in order to capture identity-related attribute features, and optimizes for location and weather prediction to capture additional facial attribute features. Finally, the network is fine-tuned with manually annotated samples. We perform an extensive experimental analysis on wearable data and two standard benchmark datasets based on web images (LFWA and CelebA). Our method outperforms by a large margin a network trained from scratch. Moreover, even without using manually annotated identity labels for pre-training as in previous methods, our approach achieves results that are better than the state of the art.",
"We propose a novel unsupervised learning approach to build features suitable for object detection and classification. The features are pre-trained on a large dataset without human annotation and later transferred via fine-tuning on a different, smaller and labeled dataset. The pre-training consists of solving jigsaw puzzles of natural images. To facilitate the transfer of features to other tasks, we introduce the context-free network (CFN), a siamese-ennead convolutional neural network. The features correspond to the columns of the CFN and they process image tiles independently (i.e., free of context). The later layers of the CFN then use the features to identify their geometric arrangement. Our experimental evaluations show that the learned features capture semantically relevant content. We pre-train the CFN on the training set of the ILSVRC2012 dataset and transfer the features on the combined training and validation set of Pascal VOC 2007 for object detection (via fast RCNN) and classification. These features outperform all current unsupervised features with (51.8 , ) for detection and (68.6 , ) for classification, and reduce the gap with supervised learning ( (56.5 , ) and (78.2 , ) respectively).",
"We learn rich natural sound representations by capitalizing on large amounts of unlabeled sound data collected in the wild. We leverage the natural synchronization between vision and sound to learn an acoustic representation using two-million unlabeled videos. Unlabeled video has the advantage that it can be economically acquired at massive scales, yet contains useful signals about natural sound. We propose a student-teacher training procedure which transfers discriminative visual knowledge from well established visual recognition models into the sound modality using unlabeled video as a bridge. Our sound representation yields significant performance improvements over the state-of-the-art results on standard benchmarks for acoustic scene object classification. Visualizations suggest some high-level semantics automatically emerge in the sound network, even though it is trained without ground truth labels.",
"Understanding how images of objects and scenes behave in response to specific ego-motions is a crucial aspect of proper visual development, yet existing visual learning methods are conspicuously disconnected from the physical source of their images. We propose to exploit proprioceptive motor signals to provide unsupervised regularization in convolutional neural networks to learn visual representations from egocentric video. Specifically, we enforce that our learned features exhibit equivariance, i.e, they respond predictably to transformations associated with distinct ego-motions. With three datasets, we show that our unsupervised feature learning approach significantly outperforms previous approaches on visual recognition and next-best-view prediction tasks. In the most challenging test, we show that features learned from video captured on an autonomous driving platform improve large-scale scene recognition in static images from a disjoint domain.",
"We present an unsupervised visual feature learning algorithm driven by context-based pixel prediction. By analogy with auto-encoders, we propose Context Encoders -- a convolutional neural network trained to generate the contents of an arbitrary image region conditioned on its surroundings. In order to succeed at this task, context encoders need to both understand the content of the entire image, as well as produce a plausible hypothesis for the missing part(s). When training context encoders, we have experimented with both a standard pixel-wise reconstruction loss, as well as a reconstruction plus an adversarial loss. The latter produces much sharper results because it can better handle multiple modes in the output. We found that a context encoder learns a representation that captures not just appearance but also the semantics of visual structures. We quantitatively demonstrate the effectiveness of our learned features for CNN pre-training on classification, detection, and segmentation tasks. Furthermore, context encoders can be used for semantic inpainting tasks, either stand-alone or as initialization for non-parametric methods.",
"This work proposes a learning method for deep architectures that takes advantage of sequential data, in particular from the temporal coherence that naturally exists in unlabeled video recordings. That is, two successive frames are likely to contain the same object or objects. This coherence is used as a supervisory signal over the unlabeled data, and is used to improve the performance on a supervised task of interest. We demonstrate the effectiveness of this method on some pose invariant object and face recognition tasks.",
"Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result. The system is implemented as a feed-forward pass in a CNN at test time and is trained on over a million color images. We evaluate our algorithm using a “colorization Turing test,” asking human participants to choose between a generated and ground truth color image. Our method successfully fools humans on 32 of the trials, significantly higher than previous methods. Moreover, we show that colorization can be a powerful pretext task for self-supervised feature learning, acting as a cross-channel encoder. This approach results in state-of-the-art performance on several feature learning benchmarks.",
"The dominant paradigm for feature learning in computer vision relies on training neural networks for the task of object recognition using millions of hand labelled images. Is it possible to learn useful features for a diverse set of visual tasks using any other form of supervision? In biology, living organisms developed the ability of visual perception for the purpose of moving and acting in the world. Drawing inspiration from this observation, in this work we investigate if the awareness of egomotion can be used as a supervisory signal for feature learning. As opposed to the knowledge of class labels, information about egomotion is freely available to mobile agents. We show that given the same number of training images, features learnt using egomotion as supervision compare favourably to features learnt using class-label as supervision on visual tasks of scene recognition, object recognition, visual odometry and keypoint matching."
]
} |
1707.06971 | 2739007515 | We propose a new sentence simplification task (Split-and-Rephrase) where the aim is to split a complex sentence into a meaning preserving sequence of shorter sentences. Like sentence simplification, splitting-and-rephrasing has the potential of benefiting both natural language processing and societ al applications. Because shorter sentences are generally better processed by NLP systems, it could be used as a preprocessing step which facilitates and improves the performance of parsers, semantic role labellers and machine translation systems. It should also be of use for people with reading disabilities because it allows the conversion of longer sentences into shorter ones. This paper makes two contributions towards this new task. First, we create and make available a benchmark consisting of 1,066,115 tuples mapping a single complex sentence to a sequence of sentences expressing the same meaning. Second, we propose five models (vanilla sequence-to-sequence to semantically-motivated models) to understand the difficulty of the proposed task. | Paraphrasing approaches include bootstrapping approaches which start from slotted templates (e.g., X is the author of Y ) and seed (e.g., X = Jack Kerouac, Y = On the Road'' ) to iteratively learn new templates from the seeds and new seeds from the new templates @cite_9 @cite_25 ; systems which extract paraphrase patterns from large monolingual corpora and use them to rewrite an input text @cite_56 @cite_17 ; statistical machine translation (SMT) based systems which learn paraphrases from monolingual parallel @cite_50 @cite_18 , comparable @cite_32 or bilingual parallel @cite_20 @cite_19 corpora; and a recent neural machine translation (NMT) based system which learns paraphrases from bilingual parallel corpora @cite_33 . | {
"cite_N": [
"@cite_18",
"@cite_33",
"@cite_9",
"@cite_32",
"@cite_56",
"@cite_19",
"@cite_50",
"@cite_25",
"@cite_20",
"@cite_17"
],
"mid": [
"2140847984",
"2741049976",
"2167435923",
"174630521",
"2061433145",
"2130571329",
"2107130271",
"",
"2143927888",
"2255434878"
],
"abstract": [
"Paraphrase patterns are useful in paraphrase recognition and generation. In this paper, we present a pivot approach for extracting paraphrase patterns from bilingual parallel corpora, whereby the English paraphrase patterns are extracted using the sentences in a foreign language as pivots. We propose a loglinear model to compute the paraphrase likelihood of two patterns and exploit feature functions based on maximum likelihood estimation (MLE) and lexical weighting (LW). Using the presented method, we extract over 1,000,000 pairs of paraphrase patterns from 2M bilingual sentence pairs, the precision of which exceeds 67 . The evaluation results show that: (1) The pivot approach is effective in extracting paraphrase patterns, which significantly outperforms the conventional method DIRT. Especially, the log-linear model with the proposed feature functions achieves high performance. (2) The coverage of the extracted paraphrase patterns is high, which is above 84 . (3) The extracted paraphrase patterns can be classified into 5 types, which are useful in various applications.",
"",
"In this paper we explore the power of surface text patterns for open-domain question answering systems. In order to obtain an optimal set of patterns, we have developed a method for learning such patterns automatically. A tagged corpus is built from the Internet in a bootstrapping process by providing a few hand-crafted examples of each question type to Altavista. Patterns are then automatically extracted from the returned documents and standardized. We calculate the precision of each pattern, and the average precision for each question type. These patterns are then applied to find answers to new questions. Using the TREC-10 question set, we report results for two cases: answers determined from the TREC-10 corpus and from the web.",
"We apply statistical machine translation (SMT) tools to generate novel paraphrases of input sentences in the same language. The system is trained on large volumes of sentence pairs automatically extracted from clustered news articles available on the World Wide Web. Alignment Error Rate (AER) is measured to gauge the quality of the resulting corpus. A monotone phrasal decoder generates contextual replacements. Human evaluation shows that this system outperforms baseline paraphrase generation techniques and, in a departure from previous work, offers better coverage and scalability than the current best-of-breed paraphrasing approaches.",
"State-of-the-art Question Answering (QA) systems are very sensitive to variations in the phrasing of an information need. Finding the preferred language for such a need is a valuable task. We investigate that claim by adopting a simple MT-based paraphrasing technique and evaluating QA system performance on paraphrased questions. We found a potential increase of 35 in MRR with respect to the original question.",
"Previous work has shown that high quality phrasal paraphrases can be extracted from bilingual parallel corpora. However, it is not clear whether bitexts are an appropriate resource for extracting more sophisticated sentential paraphrases, which are more obviously learnable from monolingual parallel corpora. We extend bilingual paraphrase extraction to syntactic paraphrases and demonstrate its ability to learn a variety of general paraphrastic transformations, including passivization, dative shift, and topicalization. We discuss how our model can be adapted to many text generation tasks by augmenting its feature set, development data, and parameter estimation routine. We illustrate this adaptation by using our paraphrase model for the task of sentence compression and achieve results competitive with state-of-the-art compression systems.",
"While paraphrasing is critical both for interpretation and generation of natural language, current systems use manual or semi-automatic methods to collect paraphrases. We present an unsupervised learning algorithm for identification of paraphrases from a corpus of multiple English translations of the same source text. Our approach yields phrasal and single word lexical paraphrases as well as syntactic paraphrases.",
"",
"Previous work has used monolingual parallel corpora to extract and generate paraphrases. We show that this task can be done using bilingual parallel corpora, a much more commonly available resource. Using alignment techniques from phrase-based statistical machine translation, we show how paraphrases in one language can be identified using a phrase in another language as a pivot. We define a paraphrase probability that allows paraphrases extracted from a bilingual parallel corpus to be ranked using translation probabilities, and show how it can be refined to take contextual information into account. We evaluate our paraphrase extraction and ranking methods using a set of manual word alignments, and contrast the quality with paraphrases extracted from automatic alignments.",
"One of the limitations of semantic parsing approaches to open-domain question answering is the lexicosyntactic gap between natural language questions and knowledge base entries -- there are many ways to ask a question, all with the same answer. In this paper we propose to bridge this gap by generating paraphrases of the input question with the goal that at least one of them will be correctly mapped to a knowledge-base query. We introduce a novel grammar model for paraphrase generation that does not require any sentence-aligned paraphrase corpus. Our key idea is to leverage the flexibility and scalability of latent-variable probabilistic context-free grammars to sample paraphrases. We do an extrinsic evaluation of our paraphrases by plugging them into a semantic parser for Freebase. Our evaluation experiments on the WebQuestions benchmark dataset show that the performance of the semantic parser significantly improves over strong baselines."
]
} |
1707.06971 | 2739007515 | We propose a new sentence simplification task (Split-and-Rephrase) where the aim is to split a complex sentence into a meaning preserving sequence of shorter sentences. Like sentence simplification, splitting-and-rephrasing has the potential of benefiting both natural language processing and societ al applications. Because shorter sentences are generally better processed by NLP systems, it could be used as a preprocessing step which facilitates and improves the performance of parsers, semantic role labellers and machine translation systems. It should also be of use for people with reading disabilities because it allows the conversion of longer sentences into shorter ones. This paper makes two contributions towards this new task. First, we create and make available a benchmark consisting of 1,066,115 tuples mapping a single complex sentence to a sequence of sentences expressing the same meaning. Second, we propose five models (vanilla sequence-to-sequence to semantically-motivated models) to understand the difficulty of the proposed task. | In sentence simplification approaches, rephrasing is performed either by a machine translation @cite_7 @cite_8 @cite_31 @cite_42 @cite_29 or by a probabilistic model @cite_41 @cite_35 . Other approaches include symbolic approaches where hand-crafted rules are used e.g., to split coordinated and subordinated sentences into several, simpler clauses @cite_48 @cite_13 @cite_44 @cite_21 @cite_36 and lexical rephrasing rules are induced from the Wikipedia simplification corpus @cite_37 . | {
"cite_N": [
"@cite_35",
"@cite_37",
"@cite_7",
"@cite_8",
"@cite_41",
"@cite_48",
"@cite_29",
"@cite_42",
"@cite_21",
"@cite_36",
"@cite_44",
"@cite_31",
"@cite_13"
],
"mid": [
"2108373063",
"1872907992",
"2251831371",
"",
"",
"",
"",
"",
"2137095957",
"2251362149",
"",
"",
""
],
"abstract": [
"Text simplification aims to rewrite text into simpler versions, and thus make information accessible to a broader audience. Most previous work simplifies sentences using handcrafted rules aimed at splitting long sentences, or substitutes difficult words using a predefined dictionary. This paper presents a data-driven model based on quasi-synchronous grammar, a formalism that can naturally capture structural mismatches and complex rewrite operations. We describe how such a grammar can be induced from Wikipedia and propose an integer linear programming model for selecting the most appropriate simplification from the space of possible rewrites generated by the grammar. We show experimentally that our method creates simplifications that significantly reduce the reading difficulty of the input, while maintaining grammaticality and preserving its meaning.",
"We present an approach to text simplification based on synchronous dependency grammars. The higher level of abstraction afforded by dependency representations allows for a linguistically sound treatment of complex constructs requiring reordering and morphological change, such as conversion of passive voice to active. We present a synchronous grammar formalism in which it is easy to write rules by hand and also acquire them automatically from dependency parses of aligned English and Simple English sentences. The grammar formalism is optimised for monolingual translation in that it reuses ordering information from the source sentence where appropriate. We demonstrate the superiority of our approach over a leading contemporary system based on quasi-synchronous tree substitution grammars, both in terms of expressivity and performance.",
"In this paper we examine the sentence simplification problem as an English-to-English translation problem, utilizing a corpus of 137K aligned sentence pairs extracted by aligning English Wikipedia and Simple English Wikipedia. This data set contains the full range of transformation operations including rewording, reordering, insertion and deletion. We introduce a new translation model for text simplification that extends a phrase-based machine translation approach to include phrasal deletion. Evaluated based on three metrics that compare against a human reference (BLEU, word-F1 and SSA) our new approach performs significantly better than two text compression techniques (including T3) and the phrase-based translation system without deletion.",
"",
"",
"",
"",
"",
"We present a framework for reformulating sentences by applying transfer rules on a typed dependency representation. We specify a list of operations that the framework needs to support and argue that typed dependency structures are currently the most suitable formalism for complex lexico-syntactic paraphrasing. We demonstrate our approach by reformulating sentences expressing the discourse relation of causation using four lexico-syntactic discourse markers -- \"cause\" as a verb and as a noun, \"because\" as a conjunction and \"because of\" as a preposition.",
"We present a framework for text simplification based on applying transformation rules to a typed dependency representation produced by the Stanford parser. We test two approaches to regeneration from typed dependencies: (a) gen-light, where the transformed dependency graphs are linearised using the word order and morphology of the original sentence, with any changes coded into the transformation rules, and (b) gen-heavy, where the Stanford dependencies are reduced to a DSyntS representation and sentences are generating formally using the RealPro surface realiser. The main contribution of this paper is to compare the robustness of these approaches in the presence of parsing errors, using both a single parse and an n-best parse setting in an overgenerate and rank approach. We find that the gen-light approach is robust to parser error, particularly in the n-best parse setting. On the other hand, parsing errors cause the realiser in the gen-heavy approach to order words and phrases in ways that are disliked by our evaluators.",
"",
"",
""
]
} |
1707.06907 | 2739374478 | In this paper, we propose a multi-modal search engine for interior design that combines visual and textual queries. The goal of our engine is to retrieve interior objects, e.g. furniture or wall clocks, that share visual and aesthetic similarities with the query. Our search engine allows the user to take a photo of a room and retrieve with a high recall a list of items identical or visually similar to those present in the photo. Additionally, it allows to return other items that aesthetically and stylistically fit well together. To achieve this goal, our system blends the results obtained using textual and visual modalities. Thanks to this blending strategy, we increase the average style similarity score of the retrieved items by 11 . Our work is implemented as a Web-based application and it is planned to be opened to the public. | Traditionally, image-based search methods drew their inspiration from textual retrieval systems @cite_21 . By using @math -means clustering method in the space of local feature descriptors, such as SIFT @cite_15 , they are able to mimic textual word entities with the so-called visual words . Once the mapping from image salient keypoints to visually representative words was established, typical textual retrieval methods, such as Bag-of-Words @cite_3 could be used. Video Google @cite_2 was one of the first visual search engines that relied on this concept. Several extensions of this concept were proposed, e.g. spatial verification @cite_7 that checks for geometrical correctness of initial query and eliminates the results that are not geometrically plausible. Other descriptor pooling methods were also proposed, e.g. Fisher Vectors @cite_31 or VLAD @cite_32 . | {
"cite_N": [
"@cite_31",
"@cite_7",
"@cite_21",
"@cite_32",
"@cite_3",
"@cite_2",
"@cite_15"
],
"mid": [
"2147238549",
"2141362318",
"2128017662",
"2012592962",
"",
"2952846590",
"2151103935"
],
"abstract": [
"Within the field of pattern classification, the Fisher kernel is a powerful framework which combines the strengths of generative and discriminative approaches. The idea is to characterize a signal with a gradient vector derived from a generative probability model and to subsequently feed this representation to a discriminative classifier. We propose to apply this framework to image categorization where the input signals are images and where the underlying generative model is a visual vocabulary: a Gaussian mixture model which approximates the distribution of low-level features in images. We show that Fisher kernels can actually be understood as an extension of the popular bag-of-visterms. Our approach demonstrates excellent performance on two challenging databases: an in-house database of 19 object scene categories and the recently released VOC 2006 database. It is also very practical: it has low computational needs both at training and test time and vocabularies trained on one set of categories can be applied to another set without any significant loss in performance.",
"In this paper, we present a large-scale object retrieval system. The user supplies a query object by selecting a region of a query image, and the system returns a ranked list of images that contain the same object, retrieved from a large corpus. We demonstrate the scalability and performance of our system on a dataset of over 1 million images crawled from the photo-sharing site, Flickr [3], using Oxford landmarks as queries. Building an image-feature vocabulary is a major time and performance bottleneck, due to the size of our dataset. To address this problem we compare different scalable methods for building a vocabulary and introduce a novel quantization method based on randomized trees which we show outperforms the current state-of-the-art on an extensive ground-truth. Our experiments show that the quantization has a major effect on retrieval quality. To further improve query performance, we add an efficient spatial verification stage to re-rank the results returned from our bag-of-words model and show that this consistently improves search quality, though by less of a margin when the visual vocabulary is large. We view this work as a promising step towards much larger, \"web-scale \" image corpora.",
"A recognition scheme that scales efficiently to a large number of objects is presented. The efficiency and quality is exhibited in a live demonstration that recognizes CD-covers from a database of 40000 images of popular music CDs. The scheme builds upon popular techniques of indexing descriptors extracted from local regions, and is robust to background clutter and occlusion. The local region descriptors are hierarchically quantized in a vocabulary tree. The vocabulary tree allows a larger and more discriminatory vocabulary to be used efficiently, which we show experimentally leads to a dramatic improvement in retrieval quality. The most significant property of the scheme is that the tree directly defines the quantization. The quantization and the indexing are therefore fully integrated, essentially being one and the same. The recognition quality is evaluated through retrieval on a database with ground truth, showing the power of the vocabulary tree approach, going as high as 1 million images.",
"We address the problem of image search on a very large scale, where three constraints have to be considered jointly: the accuracy of the search, its efficiency, and the memory usage of the representation. We first propose a simple yet efficient way of aggregating local image descriptors into a vector of limited dimension, which can be viewed as a simplification of the Fisher kernel representation. We then show how to jointly optimize the dimension reduction and the indexing algorithm, so that it best preserves the quality of vector comparison. The evaluation shows that our approach significantly outperforms the state of the art: the search accuracy is comparable to the bag-of-features approach for an image representation that fits in 20 bytes. Searching a 10 million image dataset takes about 50ms.",
"",
"We demonstrate that, with the availability of distributed computation platforms such as Amazon Web Services and open-source tools, it is possible for a small engineering team to build, launch and maintain a cost-effective, large-scale visual search system with widely available tools. We also demonstrate, through a comprehensive set of live experiments at Pinterest, that content recommendation powered by visual search improve user engagement. By sharing our implementation details and the experiences learned from launching a commercial visual search engines from scratch, we hope visual search are more widely incorporated into today's commercial applications.",
"This paper presents a method for extracting distinctive invariant features from images that can be used to perform reliable matching between different views of an object or scene. The features are invariant to image scale and rotation, and are shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. The features are highly distinctive, in the sense that a single feature can be correctly matched with high probability against a large database of features from many images. This paper also describes an approach to using these features for object recognition. The recognition proceeds by matching individual features to a database of features from known objects using a fast nearest-neighbor algorithm, followed by a Hough transform to identify clusters belonging to a single object, and finally performing verification through least-squares solution for consistent pose parameters. This approach to recognition can robustly identify objects among clutter and occlusion while achieving near real-time performance."
]
} |
1707.06907 | 2739374478 | In this paper, we propose a multi-modal search engine for interior design that combines visual and textual queries. The goal of our engine is to retrieve interior objects, e.g. furniture or wall clocks, that share visual and aesthetic similarities with the query. Our search engine allows the user to take a photo of a room and retrieve with a high recall a list of items identical or visually similar to those present in the photo. Additionally, it allows to return other items that aesthetically and stylistically fit well together. To achieve this goal, our system blends the results obtained using textual and visual modalities. Thanks to this blending strategy, we increase the average style similarity score of the retrieved items by 11 . Our work is implemented as a Web-based application and it is planned to be opened to the public. | Successful applications of deep learning techniques in other computer vision applications have motivated researchers to apply those methods also to visual search. Although preliminary results did not seem promising due to lack of robustness to cropping, scaling and image clutter @cite_8 , later works proved potential of those methods in the domain of image-based retrieval. For instance, by incorporation of R-MAC technique @cite_24 image representation based on the outputs of convolutional neural networks could be computed in a fixed layout of spatial regions. Many other deep architectures were also proposed, such as siamese networks, and proved successful when applied to content-based image retrieval @cite_6 . | {
"cite_N": [
"@cite_24",
"@cite_6",
"@cite_8"
],
"mid": [
"2174726731",
"2021354639",
""
],
"abstract": [
"Recently, image representation built upon Convolutional Neural Network (CNN) has been shown to provide effective descriptors for image search, outperforming pre-CNN features as short-vector representations. Yet such models are not compatible with geometry-aware re-ranking methods and still outperformed, on some particular object retrieval benchmarks, by traditional image search systems relying on precise descriptor matching, geometric re-ranking, or query expansion. This work revisits both retrieval stages, namely initial search and re-ranking, by employing the same primitive information derived from the CNN. We build compact feature vectors that encode several image regions without the need to feed multiple inputs to the network. Furthermore, we extend integral images to handle max-pooling on convolutional layer activations, allowing us to efficiently localize matching objects. The resulting bounding box is finally used for image re-ranking. As a result, this paper significantly improves existing CNN-based recognition pipeline: We report for the first time results competing with traditional methods on the challenging Oxford5k and Paris6k datasets.",
"Popular sites like Houzz, Pinterest, and LikeThatDecor, have communities of users helping each other answer questions about products in images. In this paper we learn an embedding for visual search in interior design. Our embedding contains two different domains of product images: products cropped from internet scenes, and products in their iconic form. With such a multi-domain embedding, we demonstrate several applications of visual search including identifying products in scenes and finding stylistically similar products. To obtain the embedding, we train a convolutional neural network on pairs of images. We explore several training architectures including re-purposing object classifiers, using siamese networks, and using multitask learning. We evaluate our search quantitatively and qualitatively and demonstrate high quality results for search across multiple visual domains, enabling new applications in interior design.",
""
]
} |
1707.06907 | 2739374478 | In this paper, we propose a multi-modal search engine for interior design that combines visual and textual queries. The goal of our engine is to retrieve interior objects, e.g. furniture or wall clocks, that share visual and aesthetic similarities with the query. Our search engine allows the user to take a photo of a room and retrieve with a high recall a list of items identical or visually similar to those present in the photo. Additionally, it allows to return other items that aesthetically and stylistically fit well together. To achieve this goal, our system blends the results obtained using textual and visual modalities. Thanks to this blending strategy, we increase the average style similarity score of the retrieved items by 11 . Our work is implemented as a Web-based application and it is planned to be opened to the public. | Comparing the style similarity of two objects or scenes is one of the challenges that has to be answered when training a machine learning model for interior design retrieval application. This problem is far from being solved mainly due to the lack of a clear metric defining how to measure style similarity. Various approaches have been proposed for defining style similarity metric. Some of them focus on evaluating similarity between shapes based on their structures @cite_26 @cite_28 and measuring the differences between scales and orientations of bounding boxes. Other approach propose structure-transcending style similarity measure that accounts for element similarity, element saliency and prevalence @cite_14 . In this work, we follow @cite_1 , and define style as Nevertheless, instead of using hand-crafted features and predefined styles, we take data-driven probabilistic approach to determine stylistic similarity measure that we define in Sec. . | {
"cite_N": [
"@cite_28",
"@cite_14",
"@cite_26",
"@cite_1"
],
"mid": [
"1975807399",
"2321046130",
"2157545738",
"640900367"
],
"abstract": [
"We introduce an unsupervised co-hierarchical analysis of a set of shapes, aimed at discovering their hierarchical part structures and revealing relations between geometrically dissimilar yet functionally equivalent shape parts across the set. The core problem is that of representative co-selection. For each shape in the set, one representative hierarchy (tree) is selected from among many possible interpretations of the hierarchical structure of the shape. Collectively, the selected tree representatives maximize the within-cluster structural similarity among them. We develop an iterative algorithm for representative co-selection. At each step, a novel cluster-and-select scheme is applied to a set of candidate trees for all the shapes. The tree-to-tree distance for clustering caters to structural shape analysis by focusing on spatial arrangement of shape parts, rather than their geometric details. The final set of representative trees are unified to form a structural co-hierarchy. We demonstrate co-hierarchical analysis on families of man-made shapes exhibiting high degrees of geometric and finer-scale structural variabilities.",
"To the Editor:— I have read the editorial, \"Elements of Style,\" inThe Journal, Jan. 23, page 345. Naturally, I am flattered. I must accept the compliment with some reservation, however, because as a matter of fact I have on occasion made a number of revisions or drafts before I have been satisfied with what I have written. It seems to me that this is bound to be so.",
"We present a method for learning custom deformation handles for an object, from a co-analysis of similar objects. Our approach identifies the geometric and spatial constraints among the different parts of an object, and makes this information available through abstract shape handles. These handles allow the user to prescribe arbitrary deformation directives including free-form surface deformations. However, only a subset of admissible deformations is enabled to the user as learned from the constraint space. Example applications are presented in shape editing, co-deformation and style transfer.",
"Texts on methods with introductions Critical glossary of concepts Bibliography Index. Giorgio Vasari - lives of the artists, 1958 Carel van Mander - Dutch and Flemish Painters, 1604 Giovanni Bellori - the idea of the painter, the sculptor and the architect chosen from the higher natural beauties of nature, 1672 Johann Winckelmann - history of ancient art, 1764 Johann Wolfgang von Goethe - Von deutscher Baukunst, 1772 Jakob Burckhardt - reflections on history, 1872 Giovanni Morelli - Italian painters, 1880 Alois Tiegi: late Roman art industry, 1901 Heinrich Wolfflin - principles of art history, 1915 Paul Franki - principles of architectural history, 1914 Roger Fry - reflections, 1920 Henri Focillon - the world of forms, 1934 Alfred H. Baff - chart of modern art, 1936 Erwin Panofshky - the history of art as a humanistic discipline, 1940 Nikolaus Pevsner - an outline of European architecture, 1943 Arnold Hauser - the scope a limitations of sociology of art, 1959 Susan Sontag - against interpretation, 1964 E.H. Gombrich - in search of cultural history, 1967 William Fagg - studying primitive art, 1973 T.J. Clark - the conditions of artistic creation, 1974 John Onians - art history, Kinstegschichte and historia, 1978 Michael Baldwin, Charles Harrison and Mel Ramsden - art history, art criticism and explanation, 1981 Svetlana Alpers - interpretation without representation, or the viewing of Las Meninas, 1983 Hans Belting - the end of the history of art?, 1984 Griselda Pollock - Feminist interventions in the history of Art, 1988."
]
} |
1707.07046 | 2736486587 | With the remarkable growth of the Internet and communication technologies over the past few decades, Internet of Things (IoTs) is enabling the ubiquitous connectivity of heterogeneous physical devices with software, sensors, and actuators. IoT networks are naturally two-layer with the cloud and cellular networks coexisting with the underlaid device-todevice (D2D) communications. The connectivity of IoTs plays an important role in information dissemination for missioncritical and civilian applications. However, IoT communication networks are vulnerable to cyber attacks including the denialof- service (DoS) and jamming attacks, resulting in link removals in IoT network. In this work, we develop a heterogeneous IoT network design framework in which a network designer can add links to provide additional communication paths between two nodes or secure links against attacks by investing resources. By anticipating the strategic cyber attacks, we characterize the optimal design of secure IoT network by first providing a lower bound on the number of links a secure network requires for a given budget of protected links, and then developing a method to construct networks that satisfy the heterogeneous network design specifications. Therefore, each layer of the designed heterogeneous IoT network is resistant to a predefined level of malicious attacks with minimum resources. Finally, we provide case studies on the Internet of Battlefield Things (IoBT) to corroborate and illustrate our obtained results | Due to the increasing cyber threats, IoT security becomes a critical concern nowadays @cite_34 . Depending on the potential of cyber attackers, IoT networks face heterogeneous types of attacks @cite_5 . For example, attackers can target the edge computing nodes in IoT, e.g., RFID readers and sensor nodes. Some typical adversarial scenarios include the node replication attack by replicating one node's identification number @cite_24 , DoS by battery draining, sleep deprivation, and outage attacks @cite_19 @cite_2 . The attackers can also launch attacks through the IoT communication networks. Quintessential examples include the eavesdropping attack where the attacker captures the private information over the channel, and utilizes the information to design other tailored attacks @cite_33 . Another example is the data injection attack where the attacker can inject fraudulent packets into IoT communication links through insertion, manipulation, and replay techniques @cite_4 . In our work, we focus on the jamming and DoS attacks which lead to the link removal in IoT communication networks. | {
"cite_N": [
"@cite_4",
"@cite_33",
"@cite_24",
"@cite_19",
"@cite_2",
"@cite_5",
"@cite_34"
],
"mid": [
"2149690282",
"1565176535",
"1940397557",
"2109148997",
"1987264479",
"2508317201",
"1967561194"
],
"abstract": [
"The significant advances of hardware manufacturing technology and the development of efficient software algorithms make technically and economically feasible a network composed of numerous, small, low-cost sensors using wireless communications, that is, a wireless sensor network. WSNs have attracted intensive interest from both academia and industry due to their wide application in civil and military scenarios. In hostile scenarios, it is very important to protect WSNs from malicious attacks. Due to various resource limitations and the salient features of a wireless sensor network, the security design for such networks is significantly challenging. In this article, we present a comprehensive survey of WSN security issues that were investigated by researchers in recent years and that shed light on future directions for WSN security.",
"The Internet of Things (IoT) will feature pervasive sensing and control capabilities via a massive deployment of machine-type communication (MTC) devices. The limited hardware, low-complexity, and severe energy constraints of MTC devices present unique communication and security challenges. As a result, robust physical-layer security methods that can supplement or even replace lightweight cryptographic protocols are appealing solutions. In this paper, we present an overview of low-complexity physical-layer security schemes that are suitable for the IoT. A local IoT deployment is modeled as a composition of multiple sensor and data subnetworks, with uplink communications from sensors to controllers, and downlink communications from controllers to actuators. The state of the art in physical-layer security for sensor networks is reviewed, followed by an overview of communication network security techniques. We then pinpoint the most energy-efficient and low-complexity security techniques that are best suited for IoT sensing applications. This is followed by a discussion of candidate low-complexity schemes for communication security, such as on–off switching and space-time block codes. The paper concludes by discussing open research issues and avenues for further work, especially the need for a theoretically well-founded and holistic approach for incorporating complexity constraints in physical-layer security designs.",
"The low-cost, off-the-shelf hardware components in unshielded sensor-network nodes leave them vulnerable to compromise. With little effort, an adversary may capture nodes, analyze and replicate them, and surreptitiously insert these replicas at strategic locations within the network. Such attacks may have severe consequences; they may allow the adversary to corrupt network data or even disconnect significant parts of the network. Previous node replication detection schemes depend primarily on centralized mechanisms with single points of failure, or on neighborhood voting protocols that fail to detect distributed replications. To address these fundamental limitations, we propose two new algorithms based on emergent properties (Gligor (2004)), i.e., properties that arise only through the collective action of multiple nodes. Randomized multicast distributes node location information to randomly-selected witnesses, exploiting the birthday paradox to detect replicated nodes, while line-selected multicast uses the topology of the network to detect replication. Both algorithms provide globally-aware, distributed node-replica detection, and line-selected multicast displays particularly strong performance characteristics. We show that emergent algorithms represent a promising new approach to sensor network security; moreover, our results naturally extend to other classes of networks in which nodes can be captured, replicated and re-inserted by an adversary.",
"Developing reliable security measures against outbreaks of malware will facilitate the proliferation of wireless sensing technologies. The first step toward this goal is to investigate potential attack strategies and the extent of damage they can incur. The malware at each infective node may seek to contact more susceptible nodes by amplifying the transmission range and the media scanning rate and thereby accelerate its spread. This may however lead to (a) easier detection of the malware and thus more effective counter-measure by the network, and (b) faster depletion of the battery which may in turn thwart further spread of the infection and or exploitation of that node. We assume the viewpoint of the malware and cast the problem of dynamically selecting the transmission range and media access rate of the infective nodes as an optimal control problem. We utilize Pontryagin's maximum principle to find an optimum solution, and prove that the maximum damage can be attained using simple three-phase bang-bang strategies.",
"Ad hoc low-power wireless networks are an exciting research direction in sensing and pervasive computing. Prior security work in this area has focused primarily on denial of communication at the routing or medium access control levels. This paper explores resource depletion attacks at the routing protocol layer, which permanently disable networks by quickly draining nodes' battery power. These \"Vampire” attacks are not specific to any specific protocol, but rather rely on the properties of many popular classes of routing protocols. We find that all examined protocols are susceptible to Vampire attacks, which are devastating, difficult to detect, and are easy to carry out using as few as one malicious insider sending only protocol-compliant messages. In the worst case, a single Vampire can increase network-wide energy usage by a factor of O(N), where N in the number of network nodes. We discuss methods to mitigate these types of attacks, including a new proof-of-concept protocol that provably bounds the damage caused by Vampires during the packet forwarding phase.",
"Internet of Things (IoT), also referred to as the Internet of Objects, is envisioned as a transformative approach for providing numerous services. Compact smart devices constitute an essential part of IoT. They range widely in use, size, energy capacity, and computation power. However, the integration of these smart things into the standard Internet introduces several security challenges because the majority of Internet technologies and communication protocols were not designed to support IoT. Moreover, commercialization of IoT has led to public security concerns, including personal privacy issues, threat of cyber attacks, and organized crime. In order to provide a guideline for those who want to investigate IoT security and contribute to its improvement, this survey attempts to provide a comprehensive list of vulnerabilities and countermeasures against them on the edge-side layer of IoT, which consists of three levels: (i) edge nodes, (ii) communication, and (iii) edge computing. To achieve this goal, we first briefly describe three widely-known IoT reference models and define security in the context of IoT. Second, we discuss the possible applications of IoT and potential motivations of the attackers who target this new paradigm. Third, we discuss different attacks and threats. Fourth, we describe possible countermeasures against these attacks. Finally, we introduce two emerging security challenges not yet explained in detail in previous literature.",
"The Internet of Things, an emerging global Internet-based technical architecture facilitating the exchange of goods and services in global supply chain networks has an impact on the security and privacy of the involved stakeholders. Measures ensuring the architecture's resilience to attacks, data authentication, access control and client privacy need to be established. An adequate legal framework must take the underlying technology into account and would best be established by an international legislator, which is supplemented by the private sector according to specific needs and thereby becomes easily adjustable. The contents of the respective legislation must encompass the right to information, provisions prohibiting or restricting the use of mechanisms of the Internet of Things, rules on IT-security-legislation, provisions supporting the use of mechanisms of the Internet of Things and the establishment of a task force doing research on the legal challenges of the IoT."
]
} |
1707.07046 | 2736486587 | With the remarkable growth of the Internet and communication technologies over the past few decades, Internet of Things (IoTs) is enabling the ubiquitous connectivity of heterogeneous physical devices with software, sensors, and actuators. IoT networks are naturally two-layer with the cloud and cellular networks coexisting with the underlaid device-todevice (D2D) communications. The connectivity of IoTs plays an important role in information dissemination for missioncritical and civilian applications. However, IoT communication networks are vulnerable to cyber attacks including the denialof- service (DoS) and jamming attacks, resulting in link removals in IoT network. In this work, we develop a heterogeneous IoT network design framework in which a network designer can add links to provide additional communication paths between two nodes or secure links against attacks by investing resources. By anticipating the strategic cyber attacks, we characterize the optimal design of secure IoT network by first providing a lower bound on the number of links a secure network requires for a given budget of protected links, and then developing a method to construct networks that satisfy the heterogeneous network design specifications. Therefore, each layer of the designed heterogeneous IoT network is resistant to a predefined level of malicious attacks with minimum resources. Finally, we provide case studies on the Internet of Battlefield Things (IoBT) to corroborate and illustrate our obtained results | To mitigate the cyber threats in IoT, a large number of works have focused on addressing the security issues by using different methodologies @cite_23 . A contract-theoretic approach has been adopted to guarantee the performance of security services in the Internet of controlled things @cite_6 @cite_3 and mitigate the systemic cyber risks @cite_1 . The authors in @cite_30 have proposed a media-aware security architecture for facilitating multimedia applications in the IoT. @cite_13 has proposed a dynamic game model including pre-attack defense and post-attack recovery phases in designing resilient IoT-enabled infrastructure networks. Strategic security investment under bounded rationality in IoT has been studied in @cite_9 @cite_10 . The authors in @cite_27 have developed an interdependent strategic trust mechanism to defend against cyber attacks in IoT. | {
"cite_N": [
"@cite_30",
"@cite_10",
"@cite_9",
"@cite_1",
"@cite_3",
"@cite_6",
"@cite_27",
"@cite_23",
"@cite_13"
],
"mid": [
"2166919658",
"",
"2804615694",
"2911996279",
"2720329411",
"2528922300",
"2963077384",
"2561342496",
"2736833769"
],
"abstract": [
"An important challenge for supporting multimedia applications in the Internet of Things is the security heterogeneity of wired and wireless sensor and actuator networks. In this work, we design a new and efficient media-aware security framework for facilitating various multimedia applications in the Internet of Things. First, we present a novel multimedia traffic classification and analysis method for handling the heterogeneity of diverse applications. Then a media-aware traffic security architecture is proposed based on the given traffic classification to enable various multimedia services being available anywhere and anytime. Furthermore, we provide a design rule and strategy to achieve a good trade-off between a system?s flexibility and efficiency. To the best of our knowledge, this study is the first to provide general media-aware security architecture by jointly considering the characteristics of multimedia traffic, security service, and the Internet of Things.",
"",
"With the increasing connectivity enabled by the Internet of Things (IoT), security becomes a critical concern, and the users should invest to secure their IoT applications. Due to the massive devices in the IoT network, users cannot be aware of the security policies taken by all its connected neighbors. Instead, a user makes security decisions based on the cyber risks he perceives by observing a selected number of nodes. To this end, we propose a model which incorporates the limited attention or bounded rationality nature of players in the IoT. Specifically, each individual builds a sparse cognitive network which includes the users to respond to. Based on this simplified cognitive network representation, each user then determines his security investment policy by minimizing his own real-world security cost. The bounded rational decision-makings of players and their cognitive network formations are interdependent, and thus should be addressed in a holistic manner. We propose a Gestalt Nash equilibrium (GNE) solution concept to characterize the decisions of agents. Then, we design a proximal-based iterative algorithm to compute the GNE and show its convergence. With case studies to smart home communities, the designed algorithm can successfully identify the critical users whose decisions need to be taken into account by the other users during the security investment.",
"In this paper, we consider a delegated dynamic systemic cyber risk management problem between a resource owner (principal) and a risk manager (agent). The principal can only observe cyber risk outcomes of the network rather than the efforts that the agent spends on protecting the resources. Under this asymmetric information, the principal aims to minimize the systemic cyber risks by designing a dynamic contract that specifies the compensation flows and the anticipated efforts of the manager by taking into account his incentives and rational behaviors. We formulate a bi-level mechanism design problem for dynamic contract design which can be seen as a special class of differential game. We show that the principal has rational controllability of the systemic risk by designing an incentive compatible estimator of the agent's hidden efforts. We characterize the optimal mechanism design by reformulating the problem into a stochastic optimal control program and derive the solution explicitly. We further reveal a separation principle for dynamic risk management where the effort estimation and the compensation design can be achieved separately.",
"In this paper, we aim to establish a holistic framework that integrates the cyber-physical layers of a cloud-enabled Internet of Controlled Things (IoCT) through the lens of contract theory. At the physical layer, the device uses cloud services to operate the system. The quality of cloud services is unknown to the device, and hence the device designs a menu of contracts to enable a reliable and incentive-compatible service. Based on the received contracts, the cloud service provider (SP) serves the device by determining its optimal cyber defense strategy. A contract-based FlipCloud game is used to assess the security risk and the cloud quality of service (QoS) under advanced persistent threats. The contract design approach creates a pricing mechanism for on-demand security as a service for cloud-enabled IoCT. By focusing on high and low QoS types of cloud SPs, we find that the contract design can be divided into two regimes (regimes I and II) with respect to the provided cloud QoS. Specifically, the physical devices whose optimal contracts are in regime I always request the best possible cloud security service. In contrast, the device only asks for a cloud security level that can stabilize the system when the optimal contracts lie in regime II. We illustrate the obtained results via case studies of a cloud-enabled smart home.",
"The development of advanced wireless communication technologies and smart embedded control devices makes everything connected, leading to an emerging paradigm of the Internet of Controlled Things IoCT. IoCT consists of two layers of systems: cyber layer and physical layer. This work aims to establish a holistic framework that integrates the cyber-physical layers of the IoCT through the lens of contract theory. For the cyber layer, we use a FlipIt game to capture the cloud security. We focus on two types of cloud, high-type and low-type, in terms of their provided quality of service QoS. The cloud's type is of private information which is unknown to the contract maker. Therefore, the control system administrator CSA at the physical layer needs to design a menu of two contracts for each type of service provider SP due to this asymmetric information structure. According to the received contract, SP decides his cyber defense strategy in the FlipIt game of which the Nash equilibrium determines the QoS of the cloud, and further influences the physical system performance. The objective of CSA is to minimize the payment to the cloud SP and the control cost jointly by designing optimal contracts. Due to the interdependence between the cyber and physical layers in the cloud-enabled IoCT, we need to address the cloud security and contract design problems in an integrative manner. We find that CSA always requires the best QoS from two types of cloud. In addition, under the optimal contracts, the utilities of both SPs are constants. Furthermore, no contracts will be offered to the cloud if the resulting service cannot stabilize the physical system.",
"The cloud-enabled Internet of controlled things (IoCT) envisions a network of sensors, controllers, and actuators connected through a local cloud in order to intelligently control physical devices. Because cloud services are vulnerable to advanced persistent threats (APTs), each device in the IoCT must strategically decide whether to trust cloud services that may be compromised. In this paper, we present iSTRICT, an interdependent strategic trust mechanism for the cloud-enabled IoCT. iSTRICT is composed of three interdependent layers. In the cloud layer, iSTRICT uses FlipIt games to conceptualize APTs. In the communication layer, it captures the interaction between devices and the cloud using signaling games. In the physical layer, iSTRICT uses optimal control to quantify the utilities in the higher level games. Best response dynamics link the three layers in an overall “game-of-games,” for which the outcome is captured by a concept called Gestalt Nash equilibrium (GNE). We prove the existence of a GNE under a set of natural assumptions and develop an adaptive algorithm to iteratively compute the equilibrium. Finally, we apply iSTRICT to trust management for autonomous vehicles that rely on measurements from remote sources. We show that strategic trust in the communication layer achieves a worst-case probability of compromise for any attack and defense costs in the cyber layer.",
"",
"Infrastructure networks are vulnerable to both cyber and physical attacks. Building a secure and resilient networked system is essential for providing reliable and dependable services. To this end, we establish a two-player three-stage game framework to capture the dynamics in the infrastructure protection and recovery phases. Specifically, the goal of the infrastructure network designer is to keep the network connected before and after the attack, while the adversary aims to disconnect the network by compromising a set of links. With costs for creating and removing links, the two players aim to maximize their utilities while minimizing the costs. In this paper, we use the concept of subgame perfect equilibrium (SPE) to characterize the optimal strategies of the network defender and attacker. We derive the SPE explicitly in terms of system parameters. Finally, we use a case study of UAV-enabled communication networks for disaster recovery to corroborate the obtained analytical results."
]
} |
1707.06632 | 2738808968 | The problem of mapping human close-range proximity networks has been tackled using a variety of technical approaches. Wearable electronic devices, in particular, have proven to be particularly successful in a variety of settings relevant for research in social science, complex networks and infectious diseases dynamics. Each device and technology used for proximity sensing (e.g., RFIDs, Bluetooth, low-power radio or infrared communication, etc.) comes with specific biases on the close-range relations it records. Hence it is important to assess which statistical features of the empirical proximity networks are robust across different measurement techniques, and which modeling frameworks generalize well across empirical data. Here we compare time-resolved proximity networks recorded in different experimental settings and show that some important statistical features are robust across all settings considered. The observed universality calls for a simplified modeling approach. We show that one such simple model is indeed able to reproduce the main statistical distributions characterizing the empirical temporal networks. | The present study takes inspiration from the emerging body of work investigating the possibilities of analyzing proximity and face-to-face interactions using different kinds of wearable sensors. At present, mobile phones allow the collection of data on specific structural and temporal aspects of social interactions, offering ways to approximate social interactions as spatial proximity or as the co-location of mobile devices, e.g., by means of Bluetooth hits @cite_31 @cite_1 @cite_33 @cite_28 @cite_13 . For example, Do and Gatica Perez have proposed several topic models for capturing group interaction patterns from Bluetooth proximity networks @cite_3 @cite_16 . However, this approach does not always yield good proxies to the social interactions occurring between the individuals carrying the devices. | {
"cite_N": [
"@cite_33",
"@cite_28",
"@cite_1",
"@cite_3",
"@cite_31",
"@cite_16",
"@cite_13"
],
"mid": [
"2098166026",
"2057087916",
"2001081968",
"1999153225",
"2171436624",
"2082504357",
"1987835826"
],
"abstract": [
"We introduce the Friends and Family study, a longitudinal living laboratory in a residential community. In this study, we employ a ubiquitous computing approach, Social Functional Mechanism-design and Relationship Imaging, or Social fMRI, that combines extremely rich data collection with the ability to conduct targeted experimental interventions with study populations. We present our mobile-phone-based social and behavioral sensing system, deployed in the wild for over 15 months. Finally, we present three investigations performed during the study, looking into the connection between individuals' social behavior and their financial status, network effects in decision making, and a novel intervention aimed at increasing physical activity in the subject population. Results demonstrate the value of social factors for choice, motivation, and adherence, and enable quantifying the contribution of different incentive mechanisms.",
"Mobile phones are a pervasive platform for opportunistic sensing of behaviors and opinions. Three studies use location and communication sensors to model individual behaviors and symptoms, long-term health outcomes, and the diffusion of opinions in a community. These three analyses illustrate how mobile phones can unobtrusively monitor rich social interactions, because the underlying sensing technologies are now commonplace and readily available.",
"The co-evolution of social relationships and individual behavior in time and space has important implications, but is poorly understood because of the difficulty closely tracking the everyday life of a complete community. We offer evidence that relationships and behavior co-evolve in a student dormitory, based on monthly surveys and location tracking through resident cellular phones over a period of nine months. We demonstrate that a Markov jump process could capture the co-evolution in terms of the rates at which residents visit places and friends.",
"Since humans are fundamentally social beings and interact frequently with others in their daily life, understanding social context is of primary importance in building context-aware applications. In this paper, using smartphone Bluetooth as a proximity sensor to create social networks, we present a probabilistic approach to mine human interaction types in real life. Our analysis is conducted on Bluetooth data continuously sensed with smartphones for over one year from 40 individuals who are professionally or personally related. The results show that the model can automatically discover a variety of social contexts. We objectively validated our model by studying its predictive and retrieval performance.",
"An important question in behavioral epidemiology and public health is to understand how individual behavior is affected by illness and stress. Although changes in individual behavior are intertwined with contagion, epidemiologists today do not have sensing or modeling tools to quantitatively measure its effects in real-world conditions. In this paper, we propose a novel application of ubiquitous computing. We use mobile phone based co-location and communication sensing to measure characteristic behavior changes in symptomatic individuals, reflected in their total communication, interactions with respect to time of day (e.g., late night, early morning), diversity and entropy of face-to-face interactions and movement. Using these extracted mobile features, it is possible to predict the health status of an individual, without having actual health measurements from the subject. Finally, we estimate the temporal information flux and implied causality between physical symptoms, behavior and mental health.",
"While our daily activities usually involve interactions with others, the current methods on activity recognition do not often exploit the relationship between social interactions and human activity. This paper addresses the problem of interpreting social activity from human interactions captured by mobile sensing networks. Our first goal is to discover different social activities such as chatting with friends from interaction logs and then characterize them by the set of people involved, and the time and location of the occurring event. Our second goal is to perform automatic labeling of the discovered activities using predefined semantic labels such as coffee breaks, weekly meetings, or random discussions. Our analysis was conducted on a real-life interaction network sensed with Bluetooth and infrared sensors of about fifty subjects who carried sociometric badges over 6 weeks. We show that the proposed system reliably recognized coffee breaks with 99 accuracy, while weekly meetings were recognized with 88 accuracy.",
"This paper describes the deployment of a large-scale study designed to measure human interactions across a variety of communication channels, with high temporal resolution and spanning multiple years—the Copenhagen Networks Study. Specifically, we collect data on face-to-face interactions, telecommunication, social networks, location, and background information (personality, demographics, health, politics) for a densely connected population of 1 000 individuals, using state-of-the-art smartphones as social sensors. Here we provide an overview of the related work and describe the motivation and research agenda driving the study. Additionally, the paper details the data-types measured, and the technical infrastructure in terms of both backend and phone software, as well as an outline of the deployment procedures. We document the participant privacy procedures and their underlying principles. The paper is concluded with early results from data analysis, illustrating the importance of multi-channel high-resolution approach to data collection."
]
} |
1707.06632 | 2738808968 | The problem of mapping human close-range proximity networks has been tackled using a variety of technical approaches. Wearable electronic devices, in particular, have proven to be particularly successful in a variety of settings relevant for research in social science, complex networks and infectious diseases dynamics. Each device and technology used for proximity sensing (e.g., RFIDs, Bluetooth, low-power radio or infrared communication, etc.) comes with specific biases on the close-range relations it records. Hence it is important to assess which statistical features of the empirical proximity networks are robust across different measurement techniques, and which modeling frameworks generalize well across empirical data. Here we compare time-resolved proximity networks recorded in different experimental settings and show that some important statistical features are robust across all settings considered. The observed universality calls for a simplified modeling approach. We show that one such simple model is indeed able to reproduce the main statistical distributions characterizing the empirical temporal networks. | Mobile phone traces suffer a similar problem: They can be used to model human mobility @cite_23 @cite_20 with the great advantage of easily scaling up to millions of individuals; they too, however, offer only coarse localization and therefore provide only rough co-location information, yielding thus only very limited insights into the social interactions of individuals. | {
"cite_N": [
"@cite_20",
"@cite_23"
],
"mid": [
"2137558058",
"1982300822"
],
"abstract": [
"In this paper, we review some advances made recently in the study of mobile phone datasets. This area of research has emerged a decade ago, with the increasing availability of large-scale anonymized datasets, and has grown into a stand-alone topic. We survey the contributions made so far on the social networks that can be constructed with such data, the study of personal mobility, geographical partitioning, urban planning, and help towards development as well as security and privacy issues.",
"This study used a sample of 100,000 mobile phone users whose trajectory was tracked for six months to study human mobility patterns. Displacements across all users suggest behaviour close to the Levy-flight-like pattern observed previously based on the motion of marked dollar bills, but with a cutoff in the distribution. The origin of the Levy patterns observed in the aggregate data appears to be population heterogeneity and not Levy patterns at the level of the individual."
]
} |
1707.06632 | 2738808968 | The problem of mapping human close-range proximity networks has been tackled using a variety of technical approaches. Wearable electronic devices, in particular, have proven to be particularly successful in a variety of settings relevant for research in social science, complex networks and infectious diseases dynamics. Each device and technology used for proximity sensing (e.g., RFIDs, Bluetooth, low-power radio or infrared communication, etc.) comes with specific biases on the close-range relations it records. Hence it is important to assess which statistical features of the empirical proximity networks are robust across different measurement techniques, and which modeling frameworks generalize well across empirical data. Here we compare time-resolved proximity networks recorded in different experimental settings and show that some important statistical features are robust across all settings considered. The observed universality calls for a simplified modeling approach. We show that one such simple model is indeed able to reproduce the main statistical distributions characterizing the empirical temporal networks. | An alternative strategy for collecting data on social interactions is to resort to image and video processing based on cameras placed in the environment @cite_2 @cite_48 . This approach provides very rich data sets that are, in turn, computationally very complex: They require line-of-sight access to the monitored spaces and people, specific effort for equipping the relevant physical spaces, and can hardly cope with large scale data. | {
"cite_N": [
"@cite_48",
"@cite_2"
],
"mid": [
"1571434535",
"2010236945"
],
"abstract": [
"Studying free-standing conversational groups (FCGs) in unstructured social settings (e.g., cocktail party ) is gratifying due to the wealth of information available at the group (mining social networks) and individual (recognizing native behavioral and personality traits) levels. However, analyzing social scenes involving FCGs is also highly challenging due to the difficulty in extracting behavioral cues such as target locations, their speaking activity and head body pose due to crowdedness and presence of extreme occlusions. To this end, we propose SALSA , a novel dataset facilitating multimodal and Synergetic sociAL Scene Analysis, and make two main contributions to research on automated social interaction analysis: (1) SALSA records social interactions among 18 participants in a natural, indoor environment for over 60 minutes, under the poster presentation and cocktail party contexts presenting difficulties in the form of low-resolution images, lighting variations, numerous occlusions, reverberations and interfering sound sources; (2) To alleviate these problems we facilitate multimodal analysis by recording the social interplay using four static surveillance cameras and sociometric badges worn by each participant, comprising the microphone , accelerometer , bluetooth and infrared sensors. In addition to raw data, we also provide annotations concerning individuals’ personality as well as their position , head, body orientation and F-formation information over the entire event duration. Through extensive experiments with state-of-the-art approaches, we show (a) the limitations of current methods and (b) how the recorded multiple cues synergetically aid automatic analysis of social interactions. SALSA is available at h ttp: tev.fbk.eu salsa.",
"We present a novel approach for detecting social interactions in a crowded scene by employing solely visual cues. The detection of social interactions in unconstrained scenarios is a valuable and important task, especially for surveillance purposes. Our proposal is inspired by the social signaling literature, and in particular it considers the sociological notion of F-formation. An F-formation is a set of possible configurations in space that people may assume while participating in a social interaction. Our system takes as input the positions of the people in a scene and their (head) orientations; then, employing a voting strategy based on the Hough transform, it recognizes F-formations and the individuals associated with them. Experiments on simulations and real data promote our idea."
]
} |
1707.06632 | 2738808968 | The problem of mapping human close-range proximity networks has been tackled using a variety of technical approaches. Wearable electronic devices, in particular, have proven to be particularly successful in a variety of settings relevant for research in social science, complex networks and infectious diseases dynamics. Each device and technology used for proximity sensing (e.g., RFIDs, Bluetooth, low-power radio or infrared communication, etc.) comes with specific biases on the close-range relations it records. Hence it is important to assess which statistical features of the empirical proximity networks are robust across different measurement techniques, and which modeling frameworks generalize well across empirical data. Here we compare time-resolved proximity networks recorded in different experimental settings and show that some important statistical features are robust across all settings considered. The observed universality calls for a simplified modeling approach. We show that one such simple model is indeed able to reproduce the main statistical distributions characterizing the empirical temporal networks. | Since 2010, Cattuto @cite_17 have used a technique for monitoring social interactions that reconciles scalability and resolution by means of proximity-sensing systems based on active RFID devices. These devices are capable of sensing spatial proximity over different length scales and even close face-to-face interactions of individuals (1 to 2m), with tunable temporal resolution. The SocioPatterns initiative has collected and analyzed face-to-face interaction data in many different contexts. These analyses have shown strong heterogeneities in the contact duration of individuals, the robustness of these statistics across contexts, and have revealed highly non-trivial mixing patterns of individuals in schools, hospitals or offices as well as their robustness across various timescales @cite_42 @cite_50 @cite_21 @cite_47 @cite_37 . These data have been used in data-driven simulations of epidemic spreading phenomena, including the design and validation of containment measures @cite_30 . | {
"cite_N": [
"@cite_30",
"@cite_37",
"@cite_42",
"@cite_21",
"@cite_50",
"@cite_47",
"@cite_17"
],
"mid": [
"2149349114",
"2100741829",
"2166660524",
"2154656661",
"2158222386",
"2029766371",
"1995364830"
],
"abstract": [
"Background School environments are thought to play an important role in the community spread of infectious diseases such as influenza because of the high mixing rates of school children. The closure of schools has therefore been proposed as an efficient mitigation strategy. Such measures come however with high associated social and economic costs, making alternative, less disruptive interventions highly desirable. The recent availability of high-resolution contact network data from school environments provides an opportunity to design models of micro-interventions and compare the outcomes of alternative mitigation measures.",
"Empirical data on contacts between individuals in social contexts play an important role in providing information for models describing human behavior and how epidemics spread in populations. Here, we analyze data on face-to-face contacts collected in an office building. The statistical properties of contacts are similar to other social situations, but important differences are observed in the contact network structure. In particular, the contact network is strongly shaped by the organization of the offices in departments, which has consequences in the design of accurate agent-based models of epidemic spread. We consider the contact network as a potential substrate for infectious disease spread and show that its sparsity tends to prevent outbreaks of rapidly spreading epidemics. Moreover, we define three typical behaviors according to the fraction f of links each individual shares outside its own department: residents, wanderers, and linkers. Linkers ( f 50 ) act as bridges in the network and have large betweenness centralities. Thus, a vaccination strategy targeting linkers efficiently prevents large outbreaks. As such a behavior may be spotted a priori in the offices' organization or from surveys, without the full knowledge of the time-resolved contact network, this result may help the design of efficient, low-cost vaccination or social-distancing strategies.",
"Little quantitative information is available on the mixing patterns of children in school environments. Describing and understanding contacts between children at school would help quantify the transmission opportunities of respiratory infections and identify situations within schools where the risk of transmission is higher. We report on measurements carried out in a French school (6-12 years children), where we collected data on the time-resolved face-to-face proximity of children and teachers using a proximity-sensing infrastructure based on radio frequency identification devices. Data on face-to-face interactions were collected on Thursday, October 1st and Friday, October 2nd 2009. We recorded 77,602 contact events between 242 individuals (232 children and 10 teachers). In this setting, each child has on average 323 contacts per day with 47 other children, leading to an average daily interaction time of 176 minutes. Most contacts are brief, but long contacts are also observed. Contacts occur mostly within each class, and each child spends on average three times more time in contact with classmates than with children of other classes. We describe the temporal evolution of the contact network and the trajectories followed by the children in the school, which constrain the contact patterns. We determine an exposure matrix aimed at informing mathematical models. This matrix exhibits a class and age structure which is very different from the homogeneous mixing hypothesis. We report on important properties of the contact patterns between school children that are relevant for modeling the propagation of diseases and for evaluating control measures. We discuss public health implications related to the management of schools in case of epidemics and pandemics. Our results can help define a prioritization of control measures based on preventive measures, case isolation, classes and school closures, that could reduce the disruption to education during epidemics.",
"Abstract The availability of new data sources on human mobility is opening new avenues for investigating the interplay of social networks, human mobility and dynamical processes such as epidemic spreading. Here we analyze data on the time-resolved face-to-face proximity of individuals in large-scale real-world scenarios. We compare two settings with very different properties, a scientific conference and a long-running museum exhibition. We track the behavioral networks of face-to-face proximity, and characterize them from both a static and a dynamic point of view, exposing differences and similarities. We use our data to investigate the dynamics of a susceptible–infected model for epidemic spreading that unfolds on the dynamical networks of human proximity. The spreading patterns are markedly different for the conference and the museum case, and they are strongly impacted by the causal structure of the network data. A deeper study of the spreading paths shows that the mere knowledge of static aggregated networks would lead to erroneous conclusions about the transmission paths on the dynamical networks.",
"Background Nosocomial infections place a substantial burden on health care systems and represent one of the major issues in current public health, requiring notable efforts for its prevention. Understanding the dynamics of infection transmission in a hospital setting is essential for tailoring interventions and predicting the spread among individuals. Mathematical models need to be informed with accurate data on contacts among individuals. Methods and Findings We used wearable active Radio-Frequency Identification Devices (RFID) to detect face-to-face contacts among individuals with a spatial resolution of about 1.5 meters, and a time resolution of 20 seconds. The study was conducted in a general pediatrics hospital ward, during a one-week period, and included 119 participants, with 51 health care workers, 37 patients, and 31 caregivers. Nearly 16,000 contacts were recorded during the study period, with a median of approximately 20 contacts per participants per day. Overall, 25 of the contacts involved a ward assistant, 23 a nurse, 22 a patient, 22 a caregiver, and 8 a physician. The majority of contacts were of brief duration, but long and frequent contacts especially between patients and caregivers were also found. In the setting under study, caregivers do not represent a significant potential for infection spread to a large number of individuals, as their interactions mainly involve the corresponding patient. Nurses would deserve priority in prevention strategies due to their central role in the potential propagation paths of infections. Conclusions Our study shows the feasibility of accurate and reproducible measures of the pattern of contacts in a hospital setting. The obtained results are particularly useful for the study of the spread of respiratory infections, for monitoring critical patterns, and for setting up tailored prevention strategies. Proximity-sensing technology should be considered as a valuable tool for measuring such patterns and evaluating nosocomial prevention strategies in specific settings.",
"Face-to-face contacts between individuals contribute to shape social networks and play an important role in determining how infectious diseases can spread within a population. It is thus important to obtain accurate and reliable descriptions of human contact patterns occurring in various day-to-day life contexts. Recent technological advances and the development of wearable sensors able to sense proximity patterns have made it possible to gather data giving access to time-varying contact networks of individuals in specific environments. Here we present and analyze two such data sets describing with high temporal resolution the contact patterns of students in a high school. We define contact matrices describing the contact patterns between students of different classes and show the importance of the class structure. We take advantage of the fact that the two data sets were collected in the same setting during several days in two successive years to perform a longitudinal analysis on two very different timescales. We show the high stability of the contact patterns across days and across years: the statistical distributions of numbers and durations of contacts are the same in different periods, and we observe a very high similarity of the contact matrices measured in different days or different years. The rate of change of the contacts of each individual from one day to the next is also similar in different years. We discuss the interest of the present analysis and data sets for various fields, including in social sciences in order to better understand and model human behavior and interactions in different contexts, and in epidemiology in order to inform models describing the spread of infectious diseases and design targeted containment strategies.",
"Background Digital networks, mobile devices, and the possibility of mining the ever-increasing amount of digital traces that we leave behind in our daily activities are changing the way we can approach the study of human and social interactions. Large-scale datasets, however, are mostly available for collective and statistical behaviors, at coarse granularities, while high-resolution data on person-to-person interactions are generally limited to relatively small groups of individuals. Here we present a scalable experimental framework for gathering real-time data resolving face-to-face social interactions with tunable spatial and temporal granularities. Methods and Findings We use active Radio Frequency Identification (RFID) devices that assess mutual proximity in a distributed fashion by exchanging low-power radio packets. We analyze the dynamics of person-to-person interaction networks obtained in three high-resolution experiments carried out at different orders of magnitude in community size. The data sets exhibit common statistical properties and lack of a characteristic time scale from 20 seconds to several hours. The association between the number of connections and their duration shows an interesting super-linear behavior, which indicates the possibility of defining super-connectors both in the number and intensity of connections. Conclusions Taking advantage of scalability and resolution, this experimental framework allows the monitoring of social interactions, uncovering similarities in the way individuals interact in different contexts, and identifying patterns of super-connector behavior in the community. These results could impact our understanding of all phenomena driven by face-to-face interactions, such as the spreading of transmissible infectious diseases and information."
]
} |
1707.06632 | 2738808968 | The problem of mapping human close-range proximity networks has been tackled using a variety of technical approaches. Wearable electronic devices, in particular, have proven to be particularly successful in a variety of settings relevant for research in social science, complex networks and infectious diseases dynamics. Each device and technology used for proximity sensing (e.g., RFIDs, Bluetooth, low-power radio or infrared communication, etc.) comes with specific biases on the close-range relations it records. Hence it is important to assess which statistical features of the empirical proximity networks are robust across different measurement techniques, and which modeling frameworks generalize well across empirical data. Here we compare time-resolved proximity networks recorded in different experimental settings and show that some important statistical features are robust across all settings considered. The observed universality calls for a simplified modeling approach. We show that one such simple model is indeed able to reproduce the main statistical distributions characterizing the empirical temporal networks. | Along a similar line, Olguin Olguin @cite_45 have designed and employed Sociometric Badges, platforms equipped with accelerometers, microphones, Bluetooth and Infrared sensors which capture body movements, prosodic speech features, proximity and face-to-face interactions respectively. Some previous studies based on Sociometric Badges revealed important insights into human dynamics and organizational processes, such as the impact of electronic communications on the business performance of teams @cite_45 , the relationship between several behavioral features captured by Sociometric Badges, employee' self-perceptions (from surveys) and productivity @cite_45 , the spreading of personality and emotional states @cite_38 . | {
"cite_N": [
"@cite_38",
"@cite_45"
],
"mid": [
"1961143607",
"2150327087"
],
"abstract": [
"Contagion, a concept from epidemiology, has long been used to characterize social influence on people’s behavior and affective (emotional) states. While it has revealed many useful insights, it is not clear whether the contagion metaphor is sufficient to fully characterize the complex dynamics of psychological states in a social context. Using wearable sensors that capture daily face-to-face interaction, combined with three daily experience sampling surveys, we collected the most comprehensive data set of personality and emotion dynamics of an entire community of work. From this high-resolution data about actual (rather than self-reported) face-to-face interaction, a complex picture emerges where contagion (that can be seen as adaptation of behavioral responses to the behavior of other people) cannot fully capture the dynamics of transitory states. We found that social influence has two opposing effects on states: adaptation effects that go beyond mere contagion, and complementarity effects whereby individuals’ behaviors tend to complement the behaviors of others. Surprisingly, these effects can exhibit completely different directions depending on the stable personality or emotional dispositions (stable traits) of target individuals. Our findings provide a foundation for richer models of social dynamics, and have implications on organizational engineering and workplace well-being.",
"We present the design, implementation, and deployment of a wearable computing platform for measuring and analyzing human behavior in organizational settings. We propose the use of wearable electronic badges capable of automatically measuring the amount of face-to-face interaction, conversational time, physical proximity to other people, and physical activity levels in order to capture individual and collective patterns of behavior. Our goal is to be able to understand how patterns of behavior shape individuals and organizations. By using on-body sensors in large groups of people for extended periods of time in naturalistic settings, we have been able to identify, measure, and quantify social interactions, group behavior, and organizational dynamics. We deployed this wearable computing platform in a group of 22 employees working in a real organization over a period of one month. Using these automatic measurements, we were able to predict employees' self-assessments of job satisfaction and their own perceptions of group interaction quality by combining data collected with our platform and e-mail communication data. In particular, the total amount of communication was predictive of both of these assessments, and betweenness in the social network exhibited a high negative correlation with group interaction satisfaction. We also found that physical proximity and e-mail exchange had a negative correlation of r = -0.55n(p 0.01), which has far-reaching implications for past and future research on social networks."
]
} |
1707.06786 | 2737555317 | Head detection and localization is a demanding task and a key element for many computer vision applications, like video surveillance, Human Computer Interaction and face analysis. The stunning amount of work done for detecting faces on RGB images, together with the availability of huge face datasets, allowed to setup very effective systems on that domain. However, due to illumination issues, infrared or depth cameras may be required in real applications. In this paper, we introduce a novel method for head detection on depth images that exploits the classification ability of deep learning approaches. In addition to reduce the dependency on the external illumination, depth images implicitly embed useful information to deal with the scale of the target objects. Two public datasets have been exploited: the first one, called Pandora, is used to train a deep binary classifier with face and non-face images. The second one, collected by Cornell University, is used to perform a cross-dataset test during daily activities in unconstrained environments. Experimental results show that the proposed method overcomes the performance of state-of-art methods working on depth images. | Very few works present approaches for head detection only based on depth images. The recent work of Chen @cite_17 presents a novel head descriptor to classify, through a (LDA) classifier, pixels as head or non-head. Depth values are exploited to eliminate false positives of head centers and to cluster pixels for final head detection. In the work of Nghiem @cite_22 , head detection is conducted on 3D data as first step for a fall detection system. This method detects only moving objects through background subtraction and all possible head positions are searched on contour segments. Then, modified HOG features @cite_20 are computed directly on depth data, to recognize people in the image. Finally, a SVM @cite_3 is exploited to create a head shoulder classifier. Even if the presence of other recent works that exploit CNNs with depth data @cite_10 @cite_25 @cite_5 @cite_27 , we believe that this paper proposes a novel approach for head detection on only depth maps. | {
"cite_N": [
"@cite_22",
"@cite_3",
"@cite_27",
"@cite_5",
"@cite_10",
"@cite_25",
"@cite_20",
"@cite_17"
],
"mid": [
"2028456138",
"1600810180",
"2952013670",
"2741922652",
"2963670480",
"2763649054",
"2161969291",
"2549689224"
],
"abstract": [
"This article proposes a head detection algorithm for depth video provided by a Kinect camera and its application to fall detection. The proposed algorithm first detects possible head positions and then based on these positions, recognizes people by detecting the head and the shoulders. Searching for head positions is rapid because we only look for the head contour on the human outer contour. The human recognition is a modification of HOG (Histogram of Oriented Gradient) for the head and the shoulders. Compared with the original HOG, our algorithm is more robust to human articulation and back bending. The fall detection algorithm is based on the speed of the head and the body centroid and their distance to the ground. By using both the body centroid and the head, our algorithm is less affected by the centroid fluctuation. Besides, we also present a simple but effective method to verify the distance from the ground to the head and the centroid.",
"A magnetization magnet (11) is brought into contact with or close to the outer circumferential surface of a tire (12) that has a steel belt in its outer circumferential portion, and is relatively moved along the circumferential direction of the tire (12) so that the magnetic field generated by the magnetization magnet (11) runs in the circumferential direction of the tire. (12), thereby magnetizing the tire (12). The entire perimeter of the outer circumferential portion of the tire (12) is continuously magnetized in one direction along the circumferential direction. After the polarities of the magnetization magnet (11) are inverted with respect to the circumferential direction of the tire (12), an angle range smaller than 360 DEG on the outer circumferential portion of the tire is continuously re-magnetized in a direction opposite to the one direction. In a magnetic field detection method, a magnetic field is differentially detected by two magnetic detection elements, which are set so that the magnetic field detection directions of the two magnetic detection elements are parallel to the side surface of the tire, and the two elements are juxtaposed in a direction perpendicular to the side surface of the tire.",
"Recently, deep learning approaches have achieved promising results in various fields of computer vision. In this paper, we tackle the problem of head pose estimation through a Convolutional Neural Network (CNN). Differently from other proposals in the literature, the described system is able to work directly and based only on raw depth data. Moreover, the head pose estimation is solved as a regression problem and does not rely on visual facial features like facial landmarks. We tested our system on a well known public dataset, Biwi Kinect Head Pose, showing that our approach achieves state-of-art results and is able to meet real time performance requirements.",
"An accurate and fast driver's head pose estimation is a rich source of information, in particular in the automotive context. Head pose is a key element for driver's behavior investigation, pose analysis, attention monitoring and also a useful component to improve the efficacy of Human-Car Interaction systems. In this paper, a Recurrent Neural Network is exploited to tackle the problem of driver head pose estimation, directly and only working on depth images to be more reliable in presence of varying or insufficient illumination. Experimental results, obtained from two public dataset, namely Biwi Kinect Head Pose and ICT-3DHP Database, prove the efficacy of the proposed method that overcomes state-of-art works. Besides, the entire system is implemented and tested on two embedded boards with real time performance.",
"",
"A correct and reliable localization of facial landmark enables several applications in many fields, ranging from Human Computer Interaction to video surveillance. For instance, it can provide a valuable input to monitor the driver physical state and attention level in automotive context. In this paper, we tackle the problem of facial landmark localization through a deep approach. The developed system runs in real time and, in particular, is more reliable than state-of-the-art competitors specially in presence of light changes and poor illumination, thanks to the use of depth images as input. We also collected and shared a new realistic dataset inside a car, called MotorMark, to train and test the system. In addition, we exploited the public Eurecom Kinect Face Dataset for the evaluation phase, achieving promising results both in terms of accuracy and computational speed.",
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.",
"Head detection may be more demanding than face recognition and pedestrian detection in the scenarios where a face turns away or body parts are occluded in the view of a sensor, but locating people is needed. In this paper, we introduce an efficient head detection approach for single depth images at low computational expense. First, a novel head descriptor is developed and used to classify pixels as head or non-head. We use depth values to guide each window size, to eliminate false positives of head centers, and to cluster head pixels, which significantly reduce the computation costs of searching for appropriate parameters. High head detection performance was achieved in experiments − 90 accuracy for our dataset containing heads with different body postures, head poses, and distances to a Kinect2 sensor, and above 70 precision on a public dataset composed of a few daily activities, which is higher than using a head-shoulder detector with HOG feature for depth images."
]
} |
1707.06690 | 2738442461 | We study the problem of learning to reason in large scale knowledge graphs (KGs). More specifically, we describe a novel reinforcement learning framework for learning multi-hop relational paths: we use a policy-based agent with continuous states based on knowledge graph embeddings, which reasons in a KG vector space by sampling the most promising relation to extend its path. In contrast to prior work, our approach includes a reward function that takes the accuracy, diversity, and efficiency into consideration. Experimentally, we show that our proposed method outperforms a path-ranking based algorithm and knowledge graph embedding methods on Freebase and Never-Ending Language Learning datasets. | The Path-Ranking Algorithm (PRA) method @cite_16 is a primary path-finding approach that uses random walk with restart strategies for multi-hop reasoning. propose a modification to PRA that computes feature similarity in the vector space. Wang and Cohen introduce a recursive random walk approach for integrating the background KG and text---the method performs structure learning of logic programs and information extraction from text at the same time. A potential bottleneck for random walk inference is that supernodes connecting to large amount of formulas will create huge fan-out areas that significantly slow down the inference and affect the accuracy. | {
"cite_N": [
"@cite_16"
],
"mid": [
"1756422141"
],
"abstract": [
"We consider the problem of performing learning and inference in a large scale knowledge base containing imperfect knowledge with incomplete coverage. We show that a soft inference procedure based on a combination of constrained, weighted, random walks through the knowledge base graph can be used to reliably infer new beliefs for the knowledge base. More specifically, we show that the system can learn to infer different target relations by tuning the weights associated with random walks that follow different paths through the graph, using a version of the Path Ranking Algorithm (Lao and Cohen, 2010b). We apply this approach to a knowledge base of approximately 500,000 beliefs extracted imperfectly from the web by NELL, a never-ending language learner (, 2010). This new system improves significantly over NELL's earlier Horn-clause learning and inference method: it obtains nearly double the precision at rank 100, and the new learning method is also applicable to many more inference tasks."
]
} |
1707.06690 | 2738442461 | We study the problem of learning to reason in large scale knowledge graphs (KGs). More specifically, we describe a novel reinforcement learning framework for learning multi-hop relational paths: we use a policy-based agent with continuous states based on knowledge graph embeddings, which reasons in a KG vector space by sampling the most promising relation to extend its path. In contrast to prior work, our approach includes a reward function that takes the accuracy, diversity, and efficiency into consideration. Experimentally, we show that our proposed method outperforms a path-ranking based algorithm and knowledge graph embedding methods on Freebase and Never-Ending Language Learning datasets. | Neural symbolic machine @cite_17 is a more recent work on KG reasoning, which also applies reinforcement learning but has a different flavor from our work. NSM learns to compose programs that can find answers to natural language questions, while our RL model tries to add new facts to knowledge graph (KG) by reasoning on existing KG triples. In order to get answers, NSM learns to generate a sequence of actions that can be combined as a executable program. The action space in NSM is a set of predefined tokens. In our framework, the goal is to find reasoning paths, thus the action space is relation space in the KG. A similar framework @cite_9 has also been applied to visual reasoning tasks. | {
"cite_N": [
"@cite_9",
"@cite_17"
],
"mid": [
"2613404084",
"2950632879"
],
"abstract": [
"Existing methods for visual reasoning attempt to directly map inputs to outputs using black-box architectures without explicitly modeling the underlying reasoning processes. As a result, these black-box models often learn to exploit biases in the data rather than learning to perform visual reasoning. Inspired by module networks, this paper proposes a model for visual reasoning that consists of a program generator that constructs an explicit representation of the reasoning process to be performed, and an execution engine that executes the resulting program to produce an answer. Both the program generator and the execution engine are implemented by neural networks, and are trained using a combination of backpropagation and REINFORCE. Using the CLEVR benchmark for visual reasoning, we show that our model significantly outperforms strong baselines and generalizes better in a variety of settings.",
"Harnessing the statistical power of neural networks to perform language understanding and symbolic reasoning is difficult, when it requires executing efficient discrete operations against a large knowledge-base. In this work, we introduce a Neural Symbolic Machine, which contains (a) a neural \"programmer\", i.e., a sequence-to-sequence model that maps language utterances to programs and utilizes a key-variable memory to handle compositionality (b) a symbolic \"computer\", i.e., a Lisp interpreter that performs program execution, and helps find good programs by pruning the search space. We apply REINFORCE to directly optimize the task reward of this structured prediction problem. To train with weak supervision and improve the stability of REINFORCE, we augment it with an iterative maximum-likelihood training process. NSM outperforms the state-of-the-art on the WebQuestionsSP dataset when trained from question-answer pairs only, without requiring any feature engineering or domain-specific knowledge."
]
} |
1707.06665 | 2953173155 | We design and implement a distributed algorithm for balanced @math -way hypergraph partitioning that minimizes fanout, a fundamental hypergraph quantity also known as the communication volume and ( @math )-cut metric, by optimizing a novel objective called probabilistic fanout. This choice allows a simple local search heuristic to achieve comparable solution quality to the best existing hypergraph partitioners. Our algorithm is arbitrarily scalable due to a careful design that controls computational complexity, space complexity, and communication. In practice, we commonly process hypergraphs with billions of vertices and hyperedges in a few hours. We explain how the algorithm's scalability, both in terms of hypergraph size and bucket count, is limited only by the number of machines available. We perform an extensive comparison to existing distributed hypergraph partitioners and find that our approach is able to optimize hypergraphs roughly @math times bigger on the same set of machines. We call the resulting tool Social Hash Partitioner (SHP), and accompanying this paper, we open-source the most scalable version based on recursive bisection. | There exists a rich literature on graph partitioning from both theoretical and practical points of view. We refer the reader to surveys by Bichot and Siarry @cite_32 and by Bulu c et al. @cite_3 . Next we discuss existing approaches for storage sharding that utilize graph partitioning and review theoretical results on the problem. Then we describe existing libraries for hypergraph partitioning, focusing on the publicly available ones. Finally, we analyze limitations of the tools, which motivate the development of our new solution, SHP . | {
"cite_N": [
"@cite_32",
"@cite_3"
],
"mid": [
"2113513143",
"2949888389"
],
"abstract": [
"In this paper we consider the problem of (k, υ)-balanced graph partitioning - dividing the vertices of a graph into k almost equal size components (each of size less than υ • n k) so that the capacity of edges between different components is minimized. This problem is a natural generalization of several other problems such as minimum bisection, which is the (2,1)-balanced partitioning problem. We present a bicriteria polynomial time approximation algorithm with an O(log2n)-approximation for any constant υ > 1. For υ = 1 we show that no polytime approximation algorithm can guarantee a finite approximation ratio unless P=NP. Previous work has only considered the (k, υ)-balanced partitioning problem for υ ≥ 2.",
"We survey recent trends in practical algorithms for balanced graph partitioning together with applications and future research directions."
]
} |
1707.06665 | 2953173155 | We design and implement a distributed algorithm for balanced @math -way hypergraph partitioning that minimizes fanout, a fundamental hypergraph quantity also known as the communication volume and ( @math )-cut metric, by optimizing a novel objective called probabilistic fanout. This choice allows a simple local search heuristic to achieve comparable solution quality to the best existing hypergraph partitioners. Our algorithm is arbitrarily scalable due to a careful design that controls computational complexity, space complexity, and communication. In practice, we commonly process hypergraphs with billions of vertices and hyperedges in a few hours. We explain how the algorithm's scalability, both in terms of hypergraph size and bucket count, is limited only by the number of machines available. We perform an extensive comparison to existing distributed hypergraph partitioners and find that our approach is able to optimize hypergraphs roughly @math times bigger on the same set of machines. We call the resulting tool Social Hash Partitioner (SHP), and accompanying this paper, we open-source the most scalable version based on recursive bisection. | We stress that being of high theoretical importance, the above approximation algorithms are too slow to be used for large graphs, as they require solving linear programs. Hence, most of the existing methods for graph and hypergraph partitioning are heuristics based on a simple local search optimization @cite_3 . We follow the same direction and utilize a heuristic that can be implemented in an efficient way. It is unlikely that one can provide strong theoretical guarantees on a local search algorithm or improve existing bounds: Since fanout minimization is a generalization of minimum bisection, it would imply a breakthrough result. | {
"cite_N": [
"@cite_3"
],
"mid": [
"2949888389"
],
"abstract": [
"We survey recent trends in practical algorithms for balanced graph partitioning together with applications and future research directions."
]
} |
1707.06665 | 2953173155 | We design and implement a distributed algorithm for balanced @math -way hypergraph partitioning that minimizes fanout, a fundamental hypergraph quantity also known as the communication volume and ( @math )-cut metric, by optimizing a novel objective called probabilistic fanout. This choice allows a simple local search heuristic to achieve comparable solution quality to the best existing hypergraph partitioners. Our algorithm is arbitrarily scalable due to a careful design that controls computational complexity, space complexity, and communication. In practice, we commonly process hypergraphs with billions of vertices and hyperedges in a few hours. We explain how the algorithm's scalability, both in terms of hypergraph size and bucket count, is limited only by the number of machines available. We perform an extensive comparison to existing distributed hypergraph partitioners and find that our approach is able to optimize hypergraphs roughly @math times bigger on the same set of machines. We call the resulting tool Social Hash Partitioner (SHP), and accompanying this paper, we open-source the most scalable version based on recursive bisection. | * Limitations of existing solutions. Parkway and Zoltan are two hypergraph partitioners that are designed to work in a distributed environment. Both of the tools implement a multi-level coarse refine technique @cite_6 . We analyzed the algorithms and identified the following scalability limitations. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2078174680"
],
"abstract": [
"In this paper, we present a new hypergraph-partitioning algorithm that is based on the multilevel paradigm. In the multilevel paradigm, a sequence of successively coarser hypergraphs is constructed. A bisection of the smallest hypergraph is computed and it is used to obtain a bisection of the original hypergraph by successively projecting and refining the bisection to the next level finer hypergraph. We have developed new hypergraph coarsening strategies within the multilevel framework. We evaluate their performance both in terms of the size of the hyperedge cut on the bisection, as well as on the run time for a number of very large scale integration circuits. Our experiments show that our multilevel hypergraph-partitioning algorithm produces high-quality partitioning in a relatively small amount of time. The quality of the partitionings produced by our scheme are on the average 6 -23 better than those produced by other state-of-the-art schemes. Furthermore, our partitioning algorithm is significantly faster, often requiring 4-10 times less time than that required by the other schemes. Our multilevel hypergraph-partitioning algorithm scales very well for large hypergraphs. Hypergraphs with over 100 000 vertices can be bisected in a few minutes on today's workstations. Also, on the large hypergraphs, our scheme outperforms other schemes (in hyperedge cut) quite consistently with larger margins (9 -30 )."
]
} |
1707.06799 | 2738180183 | Selecting optimal parameters for a neural network architecture can often make the difference between mediocre and state-of-the-art performance. However, little is published which parameters and design choices should be evaluated or selected making the correct hyperparameter optimization often a "black art that requires expert experiences" (, 2012). In this paper, we evaluate the importance of different network design choices and hyperparameters for five common linguistic sequence tagging tasks (POS, Chunking, NER, Entity Recognition, and Event Detection). We evaluated over 50.000 different setups and found, that some parameters, like the pre-trained word embeddings or the last layer of the network, have a large impact on the performance, while other parameters, for example the number of LSTM layers or the number of recurrent units, are of minor importance. We give a recommendation on a configuration that performs well among different tasks. | Even though that it is widely recognized that the network architecture and the selected hyperparameters are crucial and adding little knobs like variational dropout @cite_17 , a CRF-classifier @cite_25 or adding character-based representations @cite_24 @cite_7 can significantly change the performance of the model, only little is reported which knobs are the most important to tune. describe a method to gain insights into the relative importance of hyperparameters by using random forest predictions with functional ANOVA compositions @cite_2 . However, it does not help to prioritize which parameter choices and extensions to implement in the first place. More practical recommendations for training deep neural network architectures and selecting hyperparameters are given by . | {
"cite_N": [
"@cite_7",
"@cite_24",
"@cite_2",
"@cite_25",
"@cite_17"
],
"mid": [
"2296283641",
"2295030615",
"2012322522",
"1940872118",
"2212703438"
],
"abstract": [
"Comunicacio presentada a la 2016 Conference of the North American Chapter of the Association for Computational Linguistics, celebrada a San Diego (CA, EUA) els dies 12 a 17 de juny 2016.",
"State-of-the-art sequence labeling systems traditionally require large amounts of task-specific knowledge in the form of hand-crafted features and data pre-processing. In this paper, we introduce a novel neutral network architecture that benefits from both word- and character-level representations automatically, by using combination of bidirectional LSTM, CNN and CRF. Our system is truly end-to-end, requiring no feature engineering or data pre-processing, thus making it applicable to a wide range of sequence labeling tasks. We evaluate our system on two data sets for two sequence labeling tasks --- Penn Treebank WSJ corpus for part-of-speech (POS) tagging and CoNLL 2003 corpus for named entity recognition (NER). We obtain state-of-the-art performance on both the two data --- 97.55 accuracy for POS tagging and 91.21 F1 for NER.",
"We investigate an application in the automatic tuning of computer codes, an area of research that has come to prominence alongside the recent rise of distributed scientific processing and heterogeneity in high-performance computing environments. Here, the response function is nonlinear and noisy and may not be smooth or stationary. Clearly needed are variable selection, decomposition of influence, and analysis of main and secondary effects for both real-valued and binary inputs and outputs. Our contribution is a novel set of tools for variable selection and sensitivity analysis based on the recently proposed dynamic tree model. We argue that this approach is uniquely well suited to the demands of our motivating example. In illustrations on benchmark data sets, we show that the new techniques are faster and offer richer feature sets than do similar approaches in the static tree and computer experiment literature. We apply the methods in code-tuning optimization, examination of a cold-cache effect, and detection of transformation errors.",
"In this paper, we propose a variety of Long Short-Term Memory (LSTM) based models for sequence tagging. These models include LSTM networks, bidirectional LSTM (BI-LSTM) networks, LSTM with a Conditional Random Field (CRF) layer (LSTM-CRF) and bidirectional LSTM with a CRF layer (BI-LSTM-CRF). Our work is the first to apply a bidirectional LSTM CRF (denoted as BI-LSTM-CRF) model to NLP benchmark sequence tagging data sets. We show that the BI-LSTM-CRF model can efficiently use both past and future input features thanks to a bidirectional LSTM component. It can also use sentence level tag information thanks to a CRF layer. The BI-LSTM-CRF model can produce state of the art (or close to) accuracy on POS, chunking and NER data sets. In addition, it is robust and has less dependence on word embedding as compared to previous observations.",
"Recurrent neural networks (RNNs) stand at the forefront of many recent developments in deep learning. Yet a major difficulty with these models is their tendency to overfit, with dropout shown to fail when applied to recurrent layers. Recent results at the intersection of Bayesian modelling and deep learning offer a Bayesian interpretation of common deep learning techniques such as dropout. This grounding of dropout in approximate Bayesian inference suggests an extension of the theoretical results, offering insights into the use of dropout with RNN models. We apply this new variational inference based dropout technique in LSTM and GRU models, assessing it on language modelling and sentiment analysis tasks. The new approach outperforms existing techniques, and to the best of our knowledge improves on the single model state-of-the-art in language modelling with the Penn Treebank (73.4 test perplexity). This extends our arsenal of variational tools in deep learning."
]
} |
1707.06813 | 2949217420 | Answer Set Programming (ASP) is a well-established formalism for nonmonotonic reasoning. An ASP program can have no answer set due to cyclic default negation. In this case, it is not possible to draw any conclusion, even if this is not intended. Recently, several paracoherent semantics have been proposed that address this issue, and several potential applications for these semantics have been identified. However, paracoherent semantics have essentially been inapplicable in practice, due to the lack of efficient algorithms and implementations. In this paper, this lack is addressed, and several different algorithms to compute semi-stable and semi-equilibrium models are proposed and implemented into an answer set solving framework. An empirical performance comparison among the new algorithms on benchmarks from ASP competitions is given as well. | Many non-monotonic semantics for logic programs with negation have been proposed that can be considered as paracoherent semantics @cite_21 @cite_25 @cite_5 @cite_29 @cite_0 @cite_31 @cite_1 @cite_20 @cite_24 . However, @cite_11 have shown that only semi-stable semantics @cite_32 and semi-equilibrium semantics @cite_11 satisfy the following desiderata properties: (i) every consistent answer set of a program corresponds to a paracoherent answer set (); (ii) if a program has some (consistent) answer set, then its paracoherent answer sets correspond to answer sets (); (iii) if a program has a classical model, then it has a paracoherent answer set (); (iv) a minimal set of atoms should be undefined (); (v) every true atom must be derived from the program (). | {
"cite_N": [
"@cite_29",
"@cite_21",
"@cite_1",
"@cite_32",
"@cite_0",
"@cite_24",
"@cite_5",
"@cite_31",
"@cite_25",
"@cite_20",
"@cite_11"
],
"mid": [
"",
"2085084839",
"",
"",
"",
"2010968193",
"1516535258",
"1567025285",
"1968513265",
"",
""
],
"abstract": [
"",
"We introduce the stable model semantics fordisjunctive logic programs and deductive databases, which generalizes the stable model semantics, defined earlier for normal (i.e., non-disjunctive) programs. Depending on whether only total (2-valued) or all partial (3-valued) models are used we obtain thedisjunctive stable semantics or thepartial disjunctive stable semantics, respectively. The proposed semantics are shown to have the following properties: • For normal programs, the disjunctive (respectively, partial disjunctive) stable semantics coincides with thestable (respectively,partial stable) semantics. • For normal programs, the partial disjunctive stable semantics also coincides with thewell-founded semantics. • For locally stratified disjunctive programs both (total and partial) disjunctive stable semantics coincide with theperfect model semantics. • The partial disjunctive stable semantics can be generalized to the class ofall disjunctive logic programs. • Both (total and partial) disjunctive stable semantics can be naturally extended to a broader class of disjunctive programs that permit the use ofclassical negation. • After translation of the programP into a suitable autoepistemic theory ( P ) the disjunctive (respectively, partial disjunctive) stable semantics ofP coincides with the autoepistemic (respectively, 3-valued autoepistemic) semantics of ( P ) .",
"",
"",
"",
"Let P be an arbitrary theory and let X be any given logic. Let M be a set of atoms. We say that M is a X-stable model of P if M is a classical model of P and P∪¬M proves in logic X all atoms in M, this is denoted by P∪¬M ⊩xM. We prove that being an X-stable model is an invariant property for disjunctive programmes under a large class of logics. Two kinds of logics are mainly considered: paraconsistent logics and normal modal logics. For modal logics we use a translation proposed by Gelfond that replaces ¬a with ¬□a. As a consequence we prove that several semantics (recently introduced) for non-monotonic reasoning are equivalent for disjunctive programmes. In addition, we show that such semantics can be characterized by a fixed-point operator in terms of classical logic. We also present a simple translation of a disjunctive programme D into a normal programme N, such that the PStable model semantics of N corresponds to the stable semantics of D over the common language. We present the formal proof of this statement.",
"Partial stable models for deductive databases, i.e., normal function-free logic programs (also called datalog programs), have two equivalent definitions: one based on 3-valued logics and another based on the notion of unfounded set. The notion of partial stable model has been extended to disjunctive deductive databases using 3-valued logics. In this paper, a characterization of partial stable models for disjunctive datalog programs is given using a suitable extension of the notion of unfounded set. Two interesting sub-classes of partial stable models, M-stable (Maximal-stable) (also called regular models, preferred extension,and maximal stable classes) and L-stable (Least undefined-stable) models, are then extended from normal to disjunctive datalog programs. On the one hand, L-stable models are shown to be the natural relaxation of the notion of total stable model; on the other hand the less strict M-stable models, endowed with a nice modularity property, may be appealing from the programming and computational point of view. M-stable and L-stable models are also compared with the regular models for disjunctive datalog programs recently proposed in the literature.",
"This paper introduces an original 2-valued semantics for Normal Logic Programs (NLP), which conservatively extends the Stable Model semantics (SM) to all normal programs. The distinction consists in the revision of one feature of SM, namely its treatment of odd loops, and of infinitely long support chains, over default negation. This single revised aspect, addressed by means of a Reductio ad Absurdum approach, affords a number of fruitful consequences, namely regarding existence, relevance and top-down querying, cumulativity, and implementation. The paper motivates and defines the Revised Stable Models semantics (rSM), justifying and exemplifying it. Properties of rSM are given and contrasted with those of SM. Furthermore, these results apply to SM whenever odd loops and infinitely long chains over negation are absent, thereby establishing significant, not previously known, properties of SM. Conclusions, further work, terminate the paper.",
"A general logic program (abbreviated to \"program\" hereafter) is a set of roles that have both positive and negative subgoals. It is common to view a deductive database as a general logic program consisting of rules (IDB) slttmg above elementary relations (EDB, facts). It is desirable to associate one Herbrand model with a program and think of that model as the \"meaning of the program, \" or Its \"declarative semantics. \" Ideally, queries directed to the program would be answered in accordance with this model. Recent research indicates that some programs do not have a \"satisfactory\" total model; for such programs, the question of an appropriate partial model arises. Unfounded sets and well-founded partial models are introduced and the well-founded semantics of a program are defined to be its well-founded partial model. If the well-founded partial model is m fact a total model. it is called the well-founded model. It n shown that the class of programs possessing a total well-founded model properly includes previously studied classes of \"stratified\" and \"locally stratified\" programs, The method in this paper is also compared with other proposals in the literature, including Clark's \"program completion, \" Fitting's and Kunen's 3-vahred interpretations of it, and the \"stable models\" of Gelfond and Lifschitz.",
"",
""
]
} |
1707.06878 | 2738090071 | Interpretability of a predictive model is a powerful feature that gains the trust of users in the correctness of the predictions. In word sense disambiguation (WSD), knowledge-based systems tend to be much more interpretable than knowledge-free counterparts as they rely on the wealth of manually-encoded elements representing word senses, such as hypernyms, usage examples, and images. We present a WSD system that bridges the gap between these two so far disconnected groups of methods. Namely, our system, providing access to several state-of-the-art WSD models, aims to be interpretable as a knowledge-based system while it remains completely unsupervised and knowledge-free. The presented tool features a Web interface for all-word disambiguation of texts that makes the sense predictions human readable by providing interpretable word sense inventories, sense representations, and disambiguation results. We provide a public API, enabling seamless integration. | DKPro WSD @cite_12 is a modular, extensible Java framework for word sense disambiguation. It implements multiple WSD methods and also provides an interface to evaluation datasets. PyWSD https: github.com alvations pywsd project also provides implementations of popular WSD methods, but these are implemented in the Python language. | {
"cite_N": [
"@cite_12"
],
"mid": [
"2132575338"
],
"abstract": [
"Implementations of word sense disambiguation (WSD) algorithms tend to be tied to a particular test corpus format and sense inventory. This makes it difficult to test their performance on new data sets, or to compare them against past algorithms implemented for different data sets. In this paper we present DKPro WSD, a freely licensed, general-purpose framework for WSD which is both modular and extensible. DKPro WSD abstracts the WSD process in such a way that test corpora, sense inventories, and algorithms can be freely swapped. Its UIMA-based architecture makes it easy to add support for new resources and algorithms. Related tasks such as word sense induction and entity linking are also supported."
]
} |
1707.06878 | 2738090071 | Interpretability of a predictive model is a powerful feature that gains the trust of users in the correctness of the predictions. In word sense disambiguation (WSD), knowledge-based systems tend to be much more interpretable than knowledge-free counterparts as they rely on the wealth of manually-encoded elements representing word senses, such as hypernyms, usage examples, and images. We present a WSD system that bridges the gap between these two so far disconnected groups of methods. Namely, our system, providing access to several state-of-the-art WSD models, aims to be interpretable as a knowledge-based system while it remains completely unsupervised and knowledge-free. The presented tool features a Web interface for all-word disambiguation of texts that makes the sense predictions human readable by providing interpretable word sense inventories, sense representations, and disambiguation results. We provide a public API, enabling seamless integration. | proposed a multi-sense extension of the Skip-gram model that features an open implementation. AdaGram @cite_8 is a system that learns sense embeddings using a Bayesian extension of the Skip-gram model and provides WSD functionality based on the induced sense inventory. SenseGram @cite_7 is a system that transforms word embeddings to sense embeddings via graph clustering and uses them for WSD. Other methods to learn sense embeddings were proposed, but these do not feature open implementations for WSD. | {
"cite_N": [
"@cite_7",
"@cite_8"
],
"mid": [
"2949364118",
"2952854841"
],
"abstract": [
"There is rising interest in vector-space word embeddings and their use in NLP, especially given recent methods for their fast estimation at very large scale. Nearly all this work, however, assumes a single vector per word type ignoring polysemy and thus jeopardizing their usefulness for downstream tasks. We present an extension to the Skip-gram model that efficiently learns multiple embeddings per word type. It differs from recent related work by jointly performing word sense discrimination and embedding learning, by non-parametrically estimating the number of senses per word type, and by its efficiency and scalability. We present new state-of-the-art results in the word similarity in context task and demonstrate its scalability by training with one machine on a corpus of nearly 1 billion tokens in less than 6 hours.",
"Recently proposed Skip-gram model is a powerful method for learning high-dimensional word representations that capture rich semantic relationships between words. However, Skip-gram as well as most prior work on learning word representations does not take into account word ambiguity and maintain only single representation per word. Although a number of Skip-gram modifications were proposed to overcome this limitation and learn multi-prototype word representations, they either require a known number of word meanings or learn them using greedy heuristic approaches. In this paper we propose the Adaptive Skip-gram model which is a nonparametric Bayesian extension of Skip-gram capable to automatically learn the required number of representations for all words at desired semantic resolution. We derive efficient online variational learning algorithm for the model and empirically demonstrate its efficiency on word-sense induction task."
]
} |
1707.06545 | 2738140362 | We present an approach to semi-supervised video object segmentation, in the context of the DAVIS 2017 challenge. Our approach combines category-based object detection, category-independent object appearance segmentation and temporal object tracking. We are motivated by the fact that the objects semantic category tends not to change throughout the video while its appearance and location can vary considerably. In order to capture the specific object appearance independent of its category, for each video we train a fully convolutional network using augmentations of the given annotated frame. We refine the appearance segmentation mask with the bounding boxes provided either by a semantic object detection network, when applicable, or by a previous frame prediction. By introducing a temporal continuity constraint on the detected boxes, we are able to improve the object segmentation mask of the appearance network and achieve competitive results on the DAVIS datasets. | Instance segmentation A closely related field is that of instance segmentation, in which the goal is to provide segmentation masks of each object instance. Several works attempt to solve this task using multi-task network cascades @cite_4 , or fully convolutional networks @cite_3 . Our method differs in that our goal is to segment an object instance for a whole video, while the object category is irrelevant. | {
"cite_N": [
"@cite_4",
"@cite_3"
],
"mid": [
"2949295283",
"2951120635"
],
"abstract": [
"Semantic segmentation research has recently witnessed rapid progress, but many leading methods are unable to identify object instances. In this paper, we present Multi-task Network Cascades for instance-aware semantic segmentation. Our model consists of three networks, respectively differentiating instances, estimating masks, and categorizing objects. These networks form a cascaded structure, and are designed to share their convolutional features. We develop an algorithm for the nontrivial end-to-end training of this causal, cascaded structure. Our solution is a clean, single-step training framework and can be generalized to cascades that have more stages. We demonstrate state-of-the-art instance-aware semantic segmentation accuracy on PASCAL VOC. Meanwhile, our method takes only 360ms testing an image using VGG-16, which is two orders of magnitude faster than previous systems for this challenging problem. As a by product, our method also achieves compelling object detection results which surpass the competitive Fast Faster R-CNN systems. The method described in this paper is the foundation of our submissions to the MS COCO 2015 segmentation competition, where we won the 1st place.",
"Fully convolutional networks (FCNs) have been proven very successful for semantic segmentation, but the FCN outputs are unaware of object instances. In this paper, we develop FCNs that are capable of proposing instance-level segment candidates. In contrast to the previous FCN that generates one score map, our FCN is designed to compute a small set of instance-sensitive score maps, each of which is the outcome of a pixel-wise classifier of a relative position to instances. On top of these instance-sensitive score maps, a simple assembling module is able to output instance candidate at each position. In contrast to the recent DeepMask method for segmenting instances, our method does not have any high-dimensional layer related to the mask resolution, but instead exploits image local coherence for estimating instances. We present competitive results of instance segment proposal on both PASCAL VOC and MS COCO."
]
} |
1707.06545 | 2738140362 | We present an approach to semi-supervised video object segmentation, in the context of the DAVIS 2017 challenge. Our approach combines category-based object detection, category-independent object appearance segmentation and temporal object tracking. We are motivated by the fact that the objects semantic category tends not to change throughout the video while its appearance and location can vary considerably. In order to capture the specific object appearance independent of its category, for each video we train a fully convolutional network using augmentations of the given annotated frame. We refine the appearance segmentation mask with the bounding boxes provided either by a semantic object detection network, when applicable, or by a previous frame prediction. By introducing a temporal continuity constraint on the detected boxes, we are able to improve the object segmentation mask of the appearance network and achieve competitive results on the DAVIS datasets. | Object detection Object detection is an important area of research in computer vision. One of the leading methods is Faster-RCNN @cite_9 . Although this method performs category-dependent detection, we are able to use in cases where an object gets a high confidence semantic detection. Along with a temporal tracking method, we show how we can enforce coherent segmentation throughout the video. | {
"cite_N": [
"@cite_9"
],
"mid": [
"2953106684"
],
"abstract": [
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available."
]
} |
1707.06776 | 2737571054 | A set of mobile robots is placed at points of an infinite line. The robots are equipped with GPS devices and they may communicate their positions on the line to a central authority. The collection contains an unknown subset of "spies", i.e., byzantine robots, which are indistinguishable from the non-faulty ones. The set of the non-faulty robots need to rendezvous in the shortest possible time in order to perform some task, while the byzantine robots may try to delay their rendezvous for as long as possible. The problem facing a central authority is to determine trajectories for all robots so as to minimize the time until the non-faulty robots have rendezvoused. The trajectories must be determined without knowledge of which robots are faulty. Our goal is to minimize the competitive ratio between the time required to achieve the first rendezvous of the non-faulty robots and the time required for such a rendezvous to occur under the assumption that the faulty robots are known at the start. We provide a bounded competitive ratio algorithm, where the central authority is informed only of the set of initial robot positions, without knowing which ones or how many of them are faulty. When an upper bound on the number of byzantine robots is known to the central authority, we provide algorithms with better competitive ratios. In some instances we are able to show these algorithms are optimal. | The mobile agent rendezvous problem has been studied extensively in many topologies (or domains) and under various assumptions on system synchronicity and capabilities of the agents @cite_19 @cite_14 @cite_21 @cite_5 both as a dynamic symmetry breaking problem @cite_6 as well as in operations research @cite_26 in order to understand the limitations of search theory. A critical distinction in the models is whether the agents must all run the same algorithm, which is generally known as the @cite_8 . If agents can execute different algorithms, generally known as the , then the problem is typically much easier, though not always trivial. | {
"cite_N": [
"@cite_14",
"@cite_26",
"@cite_8",
"@cite_21",
"@cite_6",
"@cite_19",
"@cite_5"
],
"mid": [
"2050063944",
"2040570464",
"1969585469",
"1827234103",
"",
"2092863347",
""
],
"abstract": [
"Two mobile agents (robots) having distinct labels and located in nodes of an unknown anonymous connected graph have to meet. We consider the asynchronous version of this well-studied rendezvous problem and we seek fast deterministic algorithms for it. Since in the asynchronous setting, meeting at a node, which is normally required in rendezvous, is in general impossible, we relax the demand by allowing meeting of the agents inside an edge as well. The measure of performance of a rendezvous algorithm is its cost: for a given initial location of agents in a graph, this is the number of edge traversals of both agents until rendezvous is achieved. If agents are initially situated at a distance D in an infinite line, we show a rendezvous algorithm with cost O(D|Lmin|2) when D is known and O((D + |Lmax|)3) if D is unknown, where |Lmin| and |Lmax| are the lengths of the shorter and longer label of the agents, respectively. These results still hold for the case of the ring of unknown size, but then we also give an optimal algorithm of cost O(n|Lmin|), if the size n of the ring is known, and of cost O(n|Lmax|), if it is unknown. For arbitrary graphs, we show that rendezvous is feasible if an upper bound on the size of the graph is known and we give an optimal algorithm of cost O(D|Lmin|) if the topology of the graph and the initial positions are known to agents.",
"The author considers the problem faced by two people who are placed randomly in a known search region and move about at unit speed to find each other in the least expected time. This time is called the rendezvous value of the region. It is shown how symmetries in the search region may hinder the process by preventing coordination based on concepts such as north or clockwise. A general formulation of the rendezvous search problem is given for a compact metric space endowed with a group of isometrics which represents the spatial uncertainties of the players. These concepts are illustrated by considering upper bounds for various rendezvous values for the circle and an arbitrary metric network. The discrete rendezvous problem on a cycle graph for players restricted to symmetric Markovian strategies is then solved. Finally, the author considers the problem faced by two people on an infinite line who each know the distribution of the distance but not the direction to each other.",
"The rendezvous-search problem was posed by the author 25years ago. In its basic form, it asks how two unit speed players can find each other in least expected time, when randomly placed in a known dark region. The problem received little attention until about 10 years ago. This article surveys the rapid progress that has been made since then, and also presents some new results.",
"Two mobile agents having distinct identifiers and located in nodes of an unknown anonymous connected graph, have to meet at some node of the graph. We present fast deterministic algorithms for this rendezvous problem.",
"",
"Two mobile agents, modeled as points starting at different locations of an unknown terrain, have to meet. The terrain is a polygon with polygonal holes. We consider two versions of this rendezvous problem: exact RV, when the points representing the agents have to coincide at some time, and ?-RV, when these points have to get at distance less than ? in the terrain. In any terrain, each agent chooses its trajectory, but the movements of the agent on this trajectory are controlled by an adversary that may, e.g., speed up or slow down the agent. Agents have bounded memory: their computational power is that of finite state machines. Our aim is to compare the feasibility of exact and of ?-RV when agents are anonymous vs. when they are labeled. We show classes of polygonal terrains which distinguish all the studied scenarios from the point of view of feasibility of rendezvous. The features which influence the feasibility of rendezvous include symmetries present in the terrains, boundedness of their diameter, and the number of vertices of polygons in the terrains.",
""
]
} |
1707.06776 | 2737571054 | A set of mobile robots is placed at points of an infinite line. The robots are equipped with GPS devices and they may communicate their positions on the line to a central authority. The collection contains an unknown subset of "spies", i.e., byzantine robots, which are indistinguishable from the non-faulty ones. The set of the non-faulty robots need to rendezvous in the shortest possible time in order to perform some task, while the byzantine robots may try to delay their rendezvous for as long as possible. The problem facing a central authority is to determine trajectories for all robots so as to minimize the time until the non-faulty robots have rendezvoused. The trajectories must be determined without knowledge of which robots are faulty. Our goal is to minimize the competitive ratio between the time required to achieve the first rendezvous of the non-faulty robots and the time required for such a rendezvous to occur under the assumption that the faulty robots are known at the start. We provide a bounded competitive ratio algorithm, where the central authority is informed only of the set of initial robot positions, without knowing which ones or how many of them are faulty. When an upper bound on the number of byzantine robots is known to the central authority, we provide algorithms with better competitive ratios. In some instances we are able to show these algorithms are optimal. | Closely related to our research is the work of @cite_20 and @cite_17 . In @cite_20 the authors study rendezvous of two anonymous agents, where each agent knows its own initial position in the environment, and the environment is a finite or infinite graph or a Euclidean space. They show that in the line and trees as well as in multi-dimensional Euclidean spaces and grids the agents can rendezvous in time @math , where @math is the distance between the initial positions of the agents. In @cite_17 the authors study efficient rendezvous of two mobile agents moving asynchronously in the Euclidean 2d-space. Each agent has limited visibility, permitting it to see its neighborhood at unit range from its current location. Moreover, it is assumed that each agent knows its own initial position in the plane given by its coordinates. The agents, however, are not aware of each other's position. Also worth mentioning is the work of @cite_0 which studies the rendezvous problem of location-aware agents in the asynchronous case and whose proposed algorithm provides a route, leading to rendezvous. | {
"cite_N": [
"@cite_0",
"@cite_20",
"@cite_17"
],
"mid": [
"",
"1498202041",
"1592266906"
],
"abstract": [
"",
"We study rendezvous of two anonymous agents, where each agent knows its own initial position in the environment. Their task is to meet each other as quickly as possible. The time of the rendezvous is measured by the number of synchronous rounds that agents need to use in the worst case in order to meet. In each round, an agent may make a simple move or it may stay motionless. We consider two types of environments, finite or infinite graphs and Euclidean spaces. A simple move traverses a single edge (in a graph) or at most a unit distance (in Euclidean space). The rendezvous consists in visiting by both agents the same point of the environment simultaneously (in the same round). In this paper, we propose several asymptotically optimal rendezvous algorithms. In particular, we show that in the line and trees as well as in multidimensional Euclidean spaces and grids the agents can rendezvous in time O(d), where d is the distance between the initial positions of the agents. The problem of location-aware rendezvous was studied before in the asynchronous model for Euclidean spaces and multi-dimensional grids, where the emphasis was on the length of the adopted rendezvous trajectory. We point out that, contrary to the asynchronous case, where the cost of rendezvous is dominated by the size of potentially large neighborhoods, the agents are able to meet in all graphs of at most n nodes in time almost linear in d, namely, O(d log2 n). We also determine an infinite family of graphs in which synchronized rendezvous takes time Ω(d).",
"In this paper we study efficient rendezvous of two mobile agents moving asynchronously in the Euclidean 2d-space. Each agent has limited visibility, permitting it to see its neighborhood at unit range from its current location. Moreover, it is assumed that each agent knows its own initial position in the plane given by its coordinates. The agents, however, are not aware of each others position. The agents possess coherent compasses and the same unit of length, which permit them to consider their current positions within the same system of coordinates. The cost of the rendezvous algorithm is the sum of lengths of the trajectories of both agents. This cost is taken as the maximum over all possible asynchronous movements of the agents, controlled by the adversary. We propose an algorithm that allows the agents to meet in a local neighborhood of diameter O(d), where d is the original distance between the agents. This seems rather surprising since each agent is unaware of the possible location of the other agent. In fact, the cost of our algorithm is O(d2+e), for any constant e > 0. This is almost optimal, since a lower bound of Ω(d2) is straightforward. The only up to date paper [12] on asynchronous rendezvous of bounded-visibility agents in the plane provides the feasibility proof for rendezvous, proposing a solution exponential in the distance d and in the labels of the agents. In contrast, we show here that, when the identity of the agent is based solely on its original location, an almost optimal solution is possible. An integral component of our solution is the construction of a novel type of non-simple space-filling curves that preserve locality. An infinite curve of this type visits specific grid points in the plane and provides a route that can be adopted by the mobile agents in search for one another. This new concept may also appear counter-intuitive in view of the result from [22] stating that for any simple space-filling curve, there always exists a pair of close points in the plane, such that their distance along the space-filling curve is arbitrarily large."
]
} |
1707.06776 | 2737571054 | A set of mobile robots is placed at points of an infinite line. The robots are equipped with GPS devices and they may communicate their positions on the line to a central authority. The collection contains an unknown subset of "spies", i.e., byzantine robots, which are indistinguishable from the non-faulty ones. The set of the non-faulty robots need to rendezvous in the shortest possible time in order to perform some task, while the byzantine robots may try to delay their rendezvous for as long as possible. The problem facing a central authority is to determine trajectories for all robots so as to minimize the time until the non-faulty robots have rendezvoused. The trajectories must be determined without knowledge of which robots are faulty. Our goal is to minimize the competitive ratio between the time required to achieve the first rendezvous of the non-faulty robots and the time required for such a rendezvous to occur under the assumption that the faulty robots are known at the start. We provide a bounded competitive ratio algorithm, where the central authority is informed only of the set of initial robot positions, without knowing which ones or how many of them are faulty. When an upper bound on the number of byzantine robots is known to the central authority, we provide algorithms with better competitive ratios. In some instances we are able to show these algorithms are optimal. | There is also related work on gathering a collection of identical memoryless, mobile robots in one node of an anonymous ring whereby robots start from different nodes of the ring and operate in Look-Compute-Move cycles and have to end up in the same node @cite_2 , as well as oblivious mobile robots in the same location of the plane when the robots have limited visibility @cite_24 . | {
"cite_N": [
"@cite_24",
"@cite_2"
],
"mid": [
"2010017329",
"2144182788"
],
"abstract": [
"In this paper we study the problem of gathering a collection of identical oblivious mobile robots in the same location of the plane. Previous investigations have focused mostly on the unlimited visibility setting, where each robot can always see all the others regardless of their distance.In the more difficult and realistic setting where the robots have limited visibility, the existing algorithmic results are only for convergence (towards a common point, without ever reaching it) and only for semi-synchronous environments, where robots' movements are assumed to be performed instantaneously.In contrast, we study this problem in a totally asynchronous setting, where robots' actions, computations, and movements require a finite but otherwise unpredictable amount of time. We present a protocol that allows anonymous oblivious robots with limited visibility to gather in the same location in finite time, provided they have orientation (i.e., agreement on a coordinate system).Our result indicates that, with respect to gathering, orientation is at least as powerful as instantaneous movements.",
"We consider the problem of gathering identical, memoryless, mobile robots in one node of an anonymous unoriented ring. Robots start from different nodes of the ring. They operate in Look-Compute-Move cycles and have to end up in the same node. In one cycle, a robot takes a snapshot of the current configuration (Look), makes a decision to stay idle or to move to one of its adjacent nodes (Compute), and in the latter case makes an instantaneous move to this neighbor (Move). Cycles are performed asynchronously for each robot. For an odd number of robots we prove that gathering is feasible if and only if the initial configuration is not periodic, and we provide a gathering algorithm for any such configuration. For an even number of robots we decide the feasibility of gathering except for one type of symmetric initial configurations, and provide gathering algorithms for initial configurations proved to be gatherable."
]
} |
1707.06776 | 2737571054 | A set of mobile robots is placed at points of an infinite line. The robots are equipped with GPS devices and they may communicate their positions on the line to a central authority. The collection contains an unknown subset of "spies", i.e., byzantine robots, which are indistinguishable from the non-faulty ones. The set of the non-faulty robots need to rendezvous in the shortest possible time in order to perform some task, while the byzantine robots may try to delay their rendezvous for as long as possible. The problem facing a central authority is to determine trajectories for all robots so as to minimize the time until the non-faulty robots have rendezvoused. The trajectories must be determined without knowledge of which robots are faulty. Our goal is to minimize the competitive ratio between the time required to achieve the first rendezvous of the non-faulty robots and the time required for such a rendezvous to occur under the assumption that the faulty robots are known at the start. We provide a bounded competitive ratio algorithm, where the central authority is informed only of the set of initial robot positions, without knowing which ones or how many of them are faulty. When an upper bound on the number of byzantine robots is known to the central authority, we provide algorithms with better competitive ratios. In some instances we are able to show these algorithms are optimal. | Fault tolerance has been extensively studied in distributed computing, though failures were usually related to static elements of the environment, like network nodes or links (e.g., see @cite_11 @cite_12 ), rather than to the mobile components. The unreliability of robots has been studied with respect to inaccurate robots' sensing or mobility devices (cf. @cite_23 @cite_16 @cite_13 ). Problems concerning faulty robots operating in a line environment have been studied in the context of searching in @cite_7 and patrolling @cite_22 . The questions of convergence or gathering involving faulty robots were investigated in @cite_4 @cite_15 @cite_25 @cite_9 @cite_1 . To the best of our knowledge the rendezvous problem for location aware robots some of which may be faulty has never been considered by the research community in the past. | {
"cite_N": [
"@cite_4",
"@cite_22",
"@cite_7",
"@cite_9",
"@cite_1",
"@cite_23",
"@cite_15",
"@cite_16",
"@cite_13",
"@cite_25",
"@cite_12",
"@cite_11"
],
"mid": [
"1988782771",
"",
"2477849570",
"",
"2128634904",
"2136023420",
"",
"2049232787",
"1569673105",
"2041571902",
"",
"2120510885"
],
"abstract": [
"This paper studies fault-tolerant algorithms for the problem of gathering @math autonomous mobile robots. A gathering algorithm, executed independently by each robot, must ensure that all robots are gathered at one point within finite time. In a failure-prone system, a gathering algorithm is required to successfully gather the nonfaulty robots, independently of the behavior of the faulty ones. Both crash and Byzantine faults are considered. It is first observed that most existing algorithms fail to operate correctly in a setting allowing crash failures. Subsequently, an algorithm tolerant against one crash-faulty robot in a system of three or more robots is presented. It is then observed that all known algorithms fail to operate correctly in a system prone to Byzantine faults, even in the presence of a single fault. Moreover, it is shown that in an asynchronous environment it is impossible to perform a successful gathering in a @math -robot system, even if at most one of them might fail in a Byzantine manner. Thus, the problem is studied in a fully synchronous system. An algorithm is provided in this model for gathering @math robots with at most a single faulty robot, and a more general gathering algorithm is given in an @math -robot system with up to @math faults, where @math .",
"",
"We consider the problem of searching on a line using n mobile robots, of which at most f are faulty, and the remaining are reliable. The robots start at the same location and move in parallel along the line with the same speed. There is a target placed on the line at a location unknown to the robots. Reliable robots can find the target when they reach its location, but faulty robots cannot detect the target. Our goal is to design a parallel algorithm minimizing the competitive ratio, represented by the worst case ratio between the time of arrival of the first reliable robot at the target, and the distance from the source to the target. If n ≥ 2f+2, there is a simple algorithm with competitive ratio 1. For f Our search algorithm is easily seen to be optimal for the case n=f+1. We also show that as @math tends to ∞ the competitive ratio of our algorithm for the case $n = 2f+1 approaches 3 and this is optimal. More precisely, we show that asymptotically, the competitive ratio of our proportional schedule algorithm A(2f+1,f) is at most 3 + 4ln n n , while any search algorithm has a lower bound 3 + 2ln n n on its competitive ratio.",
"",
"A team consisting of an unknown number of mobile agents, starting from different nodes of an unknown network, have to meet at the same node. Agents move in synchronous rounds. Each agent has a different label. Up to f of the agents are Byzantine. We consider two levels of Byzantine behavior. A strongly Byzantine agent can choose an arbitrary port when it moves and it can convey arbitrary information to other agents, while a weakly Byzantine agent can do the same, except changing its label. What is the minimum number of good agents that guarantees deterministic gathering of all of them, with terminationq We solve exactly this Byzantine gathering problem in arbitrary networks for weakly Byzantine agents and give approximate solutions for strongly Byzantine agents, both when the size of the network is known and when it is unknown. It turns out that both the strength versus the weakness of Byzantine behavior and the knowledge of network size significantly impact the results. For weakly Byzantine agents, we show that any number of good agents permits solving the problem for networks of known size. If the size is unknown, then this minimum number is fp2. More precisely, we show a deterministic polynomial algorithm that gathers all good agents in an arbitrary network, provided that there are at least fp2 of them. We also provide a matching lower bound: we prove that if the number of good agents is at most fp1, then they are not able to gather deterministically with termination in some networks. For strongly Byzantine agents, we give a lower bound of fp1, even when the graph is known: we show that f good agents cannot gather deterministically in the presence of f Byzantine agents even in a ring of known size. On the positive side, we give deterministic gathering algorithms for at least 2fp1 good agents when the size of the network is known and for at least 4fp2 good agents when it is unknown.",
"A number of recent studies concern algorithms for distributed control and coordination in systems of autonomous mobile robots. The common theoretical model adopted in these studies assumes that the positional input of the robots is obtained by perfectly accurate visual sensors, that robot movements are accurate, and that internal calculations performed by the robots on (real) coordinates are perfectly accurate as well. The current paper concentrates on the effect of weakening this rather strong set of assumptions and replacing it with the more realistic assumption that the robot sensors, movement, and internal calculations may have slight inaccuracies. Specifically, the paper concentrates on the ability of robot systems with inaccurate sensors, movements, and calculations to carry out the task of convergence. The paper presents several impossibility theorems, limiting the inaccuracy levels that still allow convergence, and prohibiting a general algorithm for gathering, namely, meeting at a point, in a finite number of steps. The main positive result is an algorithm for convergence under bounded measurement, movement, and calculation errors.",
"",
"Anonymous mobile robots are often classified into synchronous, semi-synchronous, and asynchronous robots when discussing the pattern formation problem. For semi-synchronous robots, all patterns formable with memory are also formable without memory, with the single exception of forming a point (i.e., the gathering) by two robots. (All patterns formable with memory are formable without memory for synchronous robots, and little is known for asynchronous robots.) However, the gathering problem for two semi-synchronous robots without memory (called oblivious robots in this paper) is trivially solvable when their local coordinate systems are consistent, and the impossibility proof essentially uses the inconsistencies in their coordinate systems. Motivated by this, this paper investigates the magnitude of consistency between the local coordinate systems necessary and sufficient to solve the gathering problem for two oblivious robots under semi-synchronous and asynchronous models. To discuss the magnitude of consistency, we assume that each robot is equipped with an unreliable compass, the bearings of which may deviate from an absolute reference direction, and that the local coordinate system of each robot is determined by its compass. We consider two families of unreliable compasses, namely, static compasses with (possibly incorrect) constant bearings and dynamic compasses the bearings of which can change arbitrarily (immediately before a new look-compute-move cycle starts and after the last cycle ends). For each of the combinations of robot and compass models, we establish the condition on deviation @math that allows an algorithm to solve the gathering problem, where the deviation is measured by the largest angle formed between the @math -axis of a compass and the reference direction of the global coordinate system: @math for semi-synchronous and asynchronous robots with static compasses, @math for semi-synchronous robots with dynamic compasses, and @math for asynchronous robots with dynamic compasses. Except for asynchronous robots with dynamic compasses, these sufficient conditions are also necessary.",
"This paper considers a system of asynchronous autonomous mobile robots that can move freely in a two-dimensional plane with no agreement on a common coordinate system. Starting from any initial configuration, the robots are required to eventually gather at a single point, not fixed in advance (gathering problem). Prior work has shown that gathering oblivious (i.e., stateless) robots cannot be achieved deterministically without additional assumptions. In particular, if robots can detect multiplicity (i.e., count robots that share the same location) gathering is possible for three or more robots. Similarly, gathering of any number of robots is possible if they share a common direction, as given by compasses, with no errors. Our work is motivated by the pragmatic standpoint that (1) compasses are error-prone devices in reality, and (2) multiplicity detection, while being easy to achieve, allows for gathering in situations with more than two robots. Consequently, this paper focusses on gathering two asynchronous mobile robots equipped with inaccurate compasses. In particular, we provide a self-stabilizing algorithm to gather, in a finite time, two oblivious robots equipped with compasses that can differ by as much as π 4.",
"This paper considers the convergence problem in autonomous mobile robot systems. A natural algorithm for the problem requires the robots to move towards their center of gravity. This paper proves the correctness of the gravitational algorithm in the fully asynchronous model. It also analyzes its convergence rate and establishes its convergence in the presence of crash faults.",
"",
"Reliable computer systems must handle malfunctioning components that give conflicting information to different parts of the system. This situation can be expressed abstractly in terms of a group of generals of the Byzantine army camped with their troops around an enemy city. Communicating only by messenger, the generals must agree upon a common battle plan. However, one or more of them may be traitors who will try to confuse the others. The problem is to find an algorithm to ensure that the loyal generals will reach agreement. It is shown that, using only oral messages, this problem is solvable if and only if more than two-thirds of the generals are loyal; so a single traitor can confound two loyal generals. With unforgeable written messages, the problem is solvable for any number of generals and possible traitors. Applications of the solutions to reliable computer systems are then discussed."
]
} |
1707.06399 | 2737853324 | Object detection aims at high speed and accuracy simultaneously. However, fast models are usually less accurate, while accurate models cannot satisfy our need for speed. A fast model can be 10 times faster but 50 less accurate than an accurate model. In this paper, we propose Adaptive Feeding (AF) to combine a fast (but less accurate) detector and an accurate (but slow) detector, by adaptively determining whether an image is easy or hard and choosing an appropriate detector for it. In practice, we build a cascade of detectors, including the AF classifier which make the easy vs. hard decision and the two detectors. The AF classifier can be tuned to obtain different tradeoff between speed and accuracy, which has negligible training time and requires no additional training data. Experimental results on the PASCAL VOC, MS COCO and Caltech Pedestrian datasets confirm that AF has the ability to achieve comparable speed as the fast detector and comparable accuracy as the accurate one at the same time. As an example, by combining the fast SSD300 with the accurate SSD500 detector, AF leads to 50 speedup over SSD500 with the same precision on the VOC2007 test set. | The deformable parts model (DPM) is a classic object detection method based on mixtures of multiscale deformable part models @cite_27 , which can capture significant variations in object appearances. It is trained using a discriminative procedure that only requires bounding boxes for the objects. DPM uses disjoint steps and histogram of gradients features @cite_1 , which is not as competitive as ConvNet-based approaches. | {
"cite_N": [
"@cite_27",
"@cite_1"
],
"mid": [
"2168356304",
"2161969291"
],
"abstract": [
"We describe an object detection system based on mixtures of multiscale deformable part models. Our system is able to represent highly variable object classes and achieves state-of-the-art results in the PASCAL object detection challenges. While deformable part models have become quite popular, their value had not been demonstrated on difficult benchmarks such as the PASCAL data sets. Our system relies on new methods for discriminative training with partially labeled data. We combine a margin-sensitive approach for data-mining hard negative examples with a formalism we call latent SVM. A latent SVM is a reformulation of MI--SVM in terms of latent variables. A latent SVM is semiconvex, and the training problem becomes convex once latent information is specified for the positive examples. This leads to an iterative training algorithm that alternates between fixing latent values for positive examples and optimizing the latent SVM objective function.",
"We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds."
]
} |
1707.06399 | 2737853324 | Object detection aims at high speed and accuracy simultaneously. However, fast models are usually less accurate, while accurate models cannot satisfy our need for speed. A fast model can be 10 times faster but 50 less accurate than an accurate model. In this paper, we propose Adaptive Feeding (AF) to combine a fast (but less accurate) detector and an accurate (but slow) detector, by adaptively determining whether an image is easy or hard and choosing an appropriate detector for it. In practice, we build a cascade of detectors, including the AF classifier which make the easy vs. hard decision and the two detectors. The AF classifier can be tuned to obtain different tradeoff between speed and accuracy, which has negligible training time and requires no additional training data. Experimental results on the PASCAL VOC, MS COCO and Caltech Pedestrian datasets confirm that AF has the ability to achieve comparable speed as the fast detector and comparable accuracy as the accurate one at the same time. As an example, by combining the fast SSD300 with the accurate SSD500 detector, AF leads to 50 speedup over SSD500 with the same precision on the VOC2007 test set. | R-CNN @cite_17 starts another revolution of object detection after DPM. R-CNN is among the first to employ deep features into detection systems, and obtained significant improvements over existing detectors at its time. However, the resulting system is very slow because features are extracted from every object proposal. Compared with R-CNN, Fast R-CNN @cite_12 not only trained the very deep VGG16 @cite_21 network but also uses ROI pooling layer @cite_14 to perform feature extraction, and was 200 @math faster at test time. After that, to speed up the proposal generation process, Faster R-CNN @cite_9 proposed the region proposal network (RPN) to generate bounding box proposals and thus achieves improvements on both speed and accuracy. Recently, ResNet @cite_16 begins to replace the VGG net in some detection systems, such as Faster R-CNN @cite_9 and R-FCN @cite_30 . However, state-of-the-art accurate detectors are in general significantly slower than real-time. | {
"cite_N": [
"@cite_30",
"@cite_14",
"@cite_9",
"@cite_21",
"@cite_16",
"@cite_12",
"@cite_17"
],
"mid": [
"2950800384",
"2109255472",
"2953106684",
"1686810756",
"2949650786",
"",
"2102605133"
],
"abstract": [
"We present region-based, fully convolutional networks for accurate and efficient object detection. In contrast to previous region-based detectors such as Fast Faster R-CNN that apply a costly per-region subnetwork hundreds of times, our region-based detector is fully convolutional with almost all computation shared on the entire image. To achieve this goal, we propose position-sensitive score maps to address a dilemma between translation-invariance in image classification and translation-variance in object detection. Our method can thus naturally adopt fully convolutional image classifier backbones, such as the latest Residual Networks (ResNets), for object detection. We show competitive results on the PASCAL VOC datasets (e.g., 83.6 mAP on the 2007 set) with the 101-layer ResNet. Meanwhile, our result is achieved at a test-time speed of 170ms per image, 2.5-20x faster than the Faster R-CNN counterpart. Code is made publicly available at: this https URL",
"Existing deep convolutional neural networks (CNNs) require a fixed-size (e.g., 224 @math 224) input image. This requirement is “artificial” and may reduce the recognition accuracy for the images or sub-images of an arbitrary size scale. In this work, we equip the networks with another pooling strategy, “spatial pyramid pooling”, to eliminate the above requirement. The new network structure, called SPP-net, can generate a fixed-length representation regardless of image size scale. Pyramid pooling is also robust to object deformations. With these advantages, SPP-net should in general improve all CNN-based image classification methods. On the ImageNet 2012 dataset, we demonstrate that SPP-net boosts the accuracy of a variety of CNN architectures despite their different designs. On the Pascal VOC 2007 and Caltech101 datasets, SPP-net achieves state-of-the-art classification results using a single full-image representation and no fine-tuning. The power of SPP-net is also significant in object detection. Using SPP-net, we compute the feature maps from the entire image only once, and then pool features in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. This method avoids repeatedly computing the convolutional features. In processing test images, our method is 24-102 @math faster than the R-CNN method, while achieving better or comparable accuracy on Pascal VOC 2007. In ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2014, our methods rank #2 in object detection and #3 in image classification among all 38 teams. This manuscript also introduces the improvement made for this competition.",
"State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.",
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
"Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57 error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28 relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.",
"",
"Object detection performance, as measured on the canonical PASCAL VOC dataset, has plateaued in the last few years. The best-performing methods are complex ensemble systems that typically combine multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30 relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3 . Our approach combines two key insights: (1) one can apply high-capacity convolutional neural networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data is scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, yields a significant performance boost. Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features. We also present experiments that provide insight into what the network learns, revealing a rich hierarchy of image features. Source code for the complete system is available at http: www.cs.berkeley.edu rbg rcnn."
]
} |
1707.06399 | 2737853324 | Object detection aims at high speed and accuracy simultaneously. However, fast models are usually less accurate, while accurate models cannot satisfy our need for speed. A fast model can be 10 times faster but 50 less accurate than an accurate model. In this paper, we propose Adaptive Feeding (AF) to combine a fast (but less accurate) detector and an accurate (but slow) detector, by adaptively determining whether an image is easy or hard and choosing an appropriate detector for it. In practice, we build a cascade of detectors, including the AF classifier which make the easy vs. hard decision and the two detectors. The AF classifier can be tuned to obtain different tradeoff between speed and accuracy, which has negligible training time and requires no additional training data. Experimental results on the PASCAL VOC, MS COCO and Caltech Pedestrian datasets confirm that AF has the ability to achieve comparable speed as the fast detector and comparable accuracy as the accurate one at the same time. As an example, by combining the fast SSD300 with the accurate SSD500 detector, AF leads to 50 speedup over SSD500 with the same precision on the VOC2007 test set. | Recently, there are also some researches utilizing cascaded and boosting methods @cite_10 @cite_18 @cite_23 @cite_15 @cite_31 . Shrivastava al @cite_10 make the traditional boosting algorithm available on deep networks which achieves higher accuracy and maintain the same detection speed. Similar to ours, Angelova al @cite_18 is based on sliding window and processes different image regions independently. However, recent deep detectors use fully-convolutional networks and take the whole image as input. On the contrary, our adaptive feeding method make a choice on the image level (not the region level) and thus saves a lot of time. | {
"cite_N": [
"@cite_18",
"@cite_15",
"@cite_23",
"@cite_31",
"@cite_10"
],
"mid": [
"2265127172",
"2474389331",
"2473640056",
"2610140147",
""
],
"abstract": [
"We present a new real-time approach to object detection that exploits the efficiency of cascade classifiers with the accuracy of deep neural networks. Deep networks have been shown to excel at classification tasks, and their ability to operate on raw pixel input without the need to design special features is very appealing. However, deep nets are notoriously slow at inference time. In this paper, we propose an approach that cascades deep nets and fast features, that is both very fast and very accurate. We apply it to the challenging task of pedestrian detection. Our algorithm runs in real-time at 15 frames per second. The resulting approach achieves a 26.2 average miss rate on the Caltech Pedestrian detection benchmark, which is competitive with the very best reported results. It is the first work we are aware of that achieves very high accuracy while running in real-time.",
"In this paper, we investigate two new strategies to detect objects accurately and efficiently using deep convolutional neural network: 1) scale-dependent pooling and 2) layerwise cascaded rejection classifiers. The scale-dependent pooling (SDP) improves detection accuracy by exploiting appropriate convolutional features depending on the scale of candidate object proposals. The cascaded rejection classifiers (CRC) effectively utilize convolutional features and eliminate negative object proposals in a cascaded manner, which greatly speeds up the detection while maintaining high accuracy. In combination of the two, our method achieves significantly better accuracy compared to other state-of-the-arts in three challenging datasets, PASCAL object detection challenge, KITTI object detection benchmark and newly collected Inner-city dataset, while being more efficient.",
"Cascade has been widely used in face detection, where classifier with low computation cost can be firstly used to shrink most of the background while keeping the recall. The cascade in detection is popularized by seminal Viola-Jones framework and then widely used in other pipelines, such as DPM and CNN. However, to our best knowledge, most of the previous detection methods use cascade in a greedy manner, where previous stages in cascade are fixed when training a new stage. So optimizations of different CNNs are isolated. In this paper, we propose joint training to achieve end-to-end optimization for CNN cascade. We show that the back propagation algorithm used in training CNN can be naturally used in training CNN cascade. We present how jointly training can be conducted on naive CNN cascade and more sophisticated region proposal network (RPN) and fast R-CNN. Experiments on face detection benchmarks verify the advantages of the joint training.",
"Deep neural networks are state of the art methods for many learning tasks due to their ability to extract increasingly better features at each network layer. However, the improved performance of additional layers in a deep network comes at the cost of added latency and energy usage in feedforward inference. As networks continue to get deeper and larger, these costs become more prohibitive for real-time and energy-sensitive applications. To address this issue, we present BranchyNet, a novel deep network architecture that is augmented with additional side branch classifiers. The architecture allows prediction results for a large portion of test samples to exit the network early via these branches when samples can already be inferred with high confidence. BranchyNet exploits the observation that features learned at an early layer of a network may often be sufficient for the classification of many data points. For more difficult samples, which are expected less frequently, BranchyNet will use further or all network layers to provide the best likelihood of correct prediction. We study the BranchyNet architecture using several well-known networks (LeNet, AlexNet, ResNet) and datasets (MNIST, CIFAR10) and show that it can both improve accuracy and significantly reduce the inference time of the network.",
""
]
} |
1707.06426 | 2736477269 | Recent development in fully convolutional neural network enables efficient end-to-end learning of semantic segmentation. Traditionally, the convolutional classifiers are taught to learn the representative semantic features of labeled semantic objects. In this work, we propose a reverse attention network (RAN) architecture that trains the network to capture the opposite concept (i.e., what are not associated with a target class) as well. The RAN is a three-branch network that performs the direct, reverse and reverse-attention learning processes simultaneously. Extensive experiments are conducted to show the effectiveness of the RAN in semantic segmentation. Being built upon the DeepLabv2-LargeFOV, the RAN achieves the state-of-the-art mIoU score (48.1 ) for the challenging PASCAL-Context dataset. Significant performance improvements are also observed for the PASCAL-VOC, Person-Part, NYUDv2 and ADE20K datasets. | A brief review on recent progresses in semantic segmentation is given in this section. Semantic segmentation is a combination of the pixel-wisea localization task @cite_4 @cite_29 and the high-level recognition task. Recent developments in deep CNNs @cite_0 @cite_16 @cite_3 enable comprehensive learning of semantic features using a large amount of image data @cite_26 @cite_21 @cite_11 . The FCN @cite_33 allows effective end-to-end learning by converting fully-connected layers into convolutional layers. | {
"cite_N": [
"@cite_26",
"@cite_4",
"@cite_33",
"@cite_29",
"@cite_21",
"@cite_3",
"@cite_0",
"@cite_16",
"@cite_11"
],
"mid": [
"2031489346",
"1979432452",
"1903029394",
"2121947440",
"",
"2950179405",
"",
"1686810756",
""
],
"abstract": [
"The Pascal Visual Object Classes (VOC) challenge is a benchmark in visual object category recognition and detection, providing the vision and machine learning communities with a standard dataset of images and annotation, and standard evaluation procedures. Organised annually from 2005 to present, the challenge and its associated dataset has become accepted as the benchmark for object detection. This paper describes the dataset and evaluation procedure. We review the state-of-the-art in evaluated methods for both classification and detection, analyse whether the methods are statistically different, what they are learning from the images (e.g. the object or its context), and what the methods find easy or confuse. The paper concludes with lessons learnt in the three year history of the challenge, and proposes directions for future improvement and extension.",
"Abstract This paper studies different methods proposed so far for segmentation evaluation. Most methods can be classified into three groups: the analytical, the empirical goodness and the empirical discrepancy groups. Each group has its own characteristics. After a brief description of each method in every group, some comparative discussions about different method groups are first carried out. An experimental comparison for some empirical (goodness and discrepancy) methods commonly used is then performed to provide a rank of their evaluation abilities. In addition, some special methods are also discussed. This study is helpful for an appropriate use of existing evaluation methods and for improving their performance as well as for systematically designing new evalution methods.",
"Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, exceed the state-of-the-art in semantic segmentation. Our key insight is to build “fully convolutional” networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning. We define and detail the space of fully convolutional networks, explain their application to spatially dense prediction tasks, and draw connections to prior models. We adapt contemporary classification networks (AlexNet [20], the VGG net [31], and GoogLeNet [32]) into fully convolutional networks and transfer their learned representations by fine-tuning [3] to the segmentation task. We then define a skip architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations. Our fully convolutional network achieves state-of-the-art segmentation of PASCAL VOC (20 relative improvement to 62.2 mean IU on 2012), NYUDv2, and SIFT Flow, while inference takes less than one fifth of a second for a typical image.",
"We propose a novel approach for solving the perceptual grouping problem in vision. Rather than focusing on local features and their consistencies in the image data, our approach aims at extracting the global impression of an image. We treat image segmentation as a graph partitioning problem and propose a novel global criterion, the normalized cut, for segmenting the graph. The normalized cut criterion measures both the total dissimilarity between the different groups as well as the total similarity within the groups. We show that an efficient computational technique based on a generalized eigenvalue problem can be used to optimize this criterion. We applied this approach to segmenting static images, as well as motion sequences, and found the results to be very encouraging.",
"",
"We propose a deep convolutional neural network architecture codenamed \"Inception\", which was responsible for setting the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC 2014). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. This was achieved by a carefully crafted design that allows for increasing the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC 2014 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.",
"",
"In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.",
""
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.