text stringlengths 17 3.36M | source stringlengths 3 333 | __index_level_0__ int64 0 518k |
|---|---|---|
The ubiquity of professional sports and specifically the NFL have lead to an increase in popularity for Fantasy Football. Users have many tools at their disposal: statistics, predictions, rankings of experts and even recommendations of peers. There are issues with all of these, though. Especially since many people pay money to play, the prediction tools should be enhanced as they provide unbiased and easy-to-use assistance for users. This paper provides and discusses approaches to predict Fantasy Football scores of Quarterbacks with relatively limited data. In addition to that, it includes several suggestions on how the data could be enhanced to achieve better results. The dataset consists only of game data from the last six NFL seasons. I used two different methods to predict the Fantasy Football scores of NFL players: Support Vector Regression (SVR) and Neural Networks. The results of both are promising given the limited data that was used. | Fantasy Football Prediction | 2,900 |
Convex potential minimisation is the de facto approach to binary classification. However, Long and Servedio [2010] proved that under symmetric label noise (SLN), minimisation of any convex potential over a linear function class can result in classification performance equivalent to random guessing. This ostensibly shows that convex losses are not SLN-robust. In this paper, we propose a convex, classification-calibrated loss and prove that it is SLN-robust. The loss avoids the Long and Servedio [2010] result by virtue of being negatively unbounded. The loss is a modification of the hinge loss, where one does not clamp at zero; hence, we call it the unhinged loss. We show that the optimal unhinged solution is equivalent to that of a strongly regularised SVM, and is the limiting solution for any convex potential; this implies that strong l2 regularisation makes most standard learners SLN-robust. Experiments confirm the SLN-robustness of the unhinged loss. | Learning with Symmetric Label Noise: The Importance of Being Unhinged | 2,901 |
A version of the dueling bandit problem is addressed in which a Condorcet winner may not exist. Two algorithms are proposed that instead seek to minimize regret with respect to the Copeland winner, which, unlike the Condorcet winner, is guaranteed to exist. The first, Copeland Confidence Bound (CCB), is designed for small numbers of arms, while the second, Scalable Copeland Bandits (SCB), works better for large-scale problems. We provide theoretical results bounding the regret accumulated by CCB and SCB, both substantially improving existing results. Such existing results either offer bounds of the form $O(K \log T)$ but require restrictive assumptions, or offer bounds of the form $O(K^2 \log T)$ without requiring such assumptions. Our results offer the best of both worlds: $O(K \log T)$ bounds without restrictive assumptions. | Copeland Dueling Bandits | 2,902 |
The outputs of a trained neural network contain much richer information than just an one-hot classifier. For example, a neural network might give an image of a dog the probability of one in a million of being a cat but it is still much larger than the probability of being a car. To reveal the hidden structure in them, we apply two unsupervised learning algorithms, PCA and ICA, to the outputs of a deep Convolutional Neural Network trained on the ImageNet of 1000 classes. The PCA/ICA embedding of the object classes reveals their visual similarity and the PCA/ICA components can be interpreted as common visual features shared by similar object classes. For an application, we proposed a new zero-shot learning method, in which the visual features learned by PCA/ICA are employed. Our zero-shot learning method achieves the state-of-the-art results on the ImageNet of over 20000 classes. | Unsupervised Learning on Neural Network Outputs: with Application in
Zero-shot Learning | 2,903 |
As we aim at alleviating the curse of high-dimensionality, subspace learning is becoming more popular. Existing approaches use either information about global or local structure of the data, and few studies simultaneously focus on global and local structures as the both of them contain important information. In this paper, we propose a global and local structure preserving sparse subspace learning (GLoSS) model for unsupervised feature selection. The model can simultaneously realize feature selection and subspace learning. In addition, we develop a greedy algorithm to establish a generic combinatorial model, and an iterative strategy based on an accelerated block coordinate descent is used to solve the GLoSS problem. We also provide whole iterate sequence convergence analysis of the proposed iterative algorithm. Extensive experiments are conducted on real-world datasets to show the superiority of the proposed approach over several state-of-the-art unsupervised feature selection approaches. | Global and Local Structure Preserving Sparse Subspace Learning: An
Iterative Approach to Unsupervised Feature Selection | 2,904 |
Biclustering involves the simultaneous clustering of objects and their attributes, thus defining local two-way clustering models. Recently, efficient algorithms were conceived to enumerate all biclusters in real-valued datasets. In this case, the solution composes a complete set of maximal and non-redundant biclusters. However, the ability to enumerate biclusters revealed a challenging scenario: in noisy datasets, each true bicluster may become highly fragmented and with a high degree of overlapping. It prevents a direct analysis of the obtained results. To revert the fragmentation, we propose here two approaches for properly aggregating the whole set of enumerated biclusters: one based on single linkage and the other directly exploring the rate of overlapping. Both proposals were compared with each other and with the actual state-of-the-art in several experiments, and they not only significantly reduced the number of biclusters but also consistently increased the quality of the solution. | On bicluster aggregation and its benefits for enumerative solutions | 2,905 |
In this paper we propose the Structured Deep Neural Network (Structured DNN) as a structured and deep learning algorithm, learning to find the best structured object (such as a label sequence) given a structured input (such as a vector sequence) by globally considering the mapping relationships between the structure rather than item by item. When automatic speech recognition is viewed as a special case of such a structured learning problem, where we have the acoustic vector sequence as the input and the phoneme label sequence as the output, it becomes possible to comprehensively learned utterance by utterance as a whole, rather than frame by frame. Structured Support Vector Machine (structured SVM) was proposed to perform ASR with structured learning previously, but limited by the linear nature of SVM. Here we propose structured DNN to use nonlinear transformations in multi-layers as a structured and deep learning algorithm. It was shown to beat structured SVM in preliminary experiments on TIMIT. | Towards Structured Deep Neural Network for Automatic Speech Recognition | 2,906 |
Unsupervised feature selection has been always attracting research attention in the communities of machine learning and data mining for decades. In this paper, we propose an unsupervised feature selection method seeking a feature coefficient matrix to select the most distinctive features. Specifically, our proposed algorithm integrates the Maximum Margin Criterion with a sparsity-based model into a joint framework, where the class margin and feature correlation are taken into account at the same time. To maximize the total data separability while preserving minimized within-class scatter simultaneously, we propose to embed Kmeans into the framework generating pseudo class label information in a scenario of unsupervised feature selection. Meanwhile, a sparsity-based model, ` 2 ,p-norm, is imposed to the regularization term to effectively discover the sparse structures of the feature coefficient matrix. In this way, noisy and irrelevant features are removed by ruling out those features whose corresponding coefficients are zeros. To alleviate the local optimum problem that is caused by random initializations of K-means, a convergence guaranteed algorithm with an updating strategy for the clustering indicator matrix, is proposed to iteractively chase the optimal solution. Performance evaluation is extensively conducted over six benchmark data sets. From plenty of experimental results, it is demonstrated that our method has superior performance against all other compared approaches. | Unsupervised Feature Analysis with Class Margin Optimization | 2,907 |
In machine learning contests such as the ImageNet Large Scale Visual Recognition Challenge and the KDD Cup, contestants can submit candidate solutions and receive from an oracle (typically the organizers of the competition) the accuracy of their guesses compared to the ground-truth labels. One of the most commonly used accuracy metrics for binary classification tasks is the Area Under the Receiver Operating Characteristics Curve (AUC). In this paper we provide proofs-of-concept of how knowledge of the AUC of a set of guesses can be used, in two different kinds of attacks, to improve the accuracy of those guesses. On the other hand, we also demonstrate the intractability of one kind of AUC exploit by proving that the number of possible binary labelings of $n$ examples for which a candidate solution obtains a AUC score of $c$ grows exponentially in $n$, for every $c\in (0,1)$. | Exploiting an Oracle that Reports AUC Scores in Machine Learning
Contests | 2,908 |
In this paper, we address the problem of multi-label classification. We consider linear classifiers and propose to learn a prior over the space of labels to directly leverage the performance of such methods. This prior takes the form of a quadratic function of the labels and permits to encode both attractive and repulsive relations between labels. We cast this problem as a structured prediction one aiming at optimizing either the accuracies of the predictors or the F 1-score. This leads to an optimization problem closely related to the max-cut problem, which naturally leads to semidefinite and spectral relaxations. We show on standard datasets how such a general prior can improve the performances of multi-label techniques. | Semidefinite and Spectral Relaxations for Multi-Label Classification | 2,909 |
Deep networks trained on large-scale data can learn transferable features to promote learning multiple tasks. Since deep features eventually transition from general to specific along deep networks, a fundamental problem of multi-task learning is how to exploit the task relatedness underlying parameter tensors and improve feature transferability in the multiple task-specific layers. This paper presents Multilinear Relationship Networks (MRN) that discover the task relationships based on novel tensor normal priors over parameter tensors of multiple task-specific layers in deep convolutional networks. By jointly learning transferable features and multilinear relationships of tasks and features, MRN is able to alleviate the dilemma of negative-transfer in the feature layers and under-transfer in the classifier layer. Experiments show that MRN yields state-of-the-art results on three multi-task learning datasets. | Learning Multiple Tasks with Multilinear Relationship Networks | 2,910 |
In this paper, we explore the inclusion of latent random variables into the dynamic hidden state of a recurrent neural network (RNN) by combining elements of the variational autoencoder. We argue that through the use of high-level latent random variables, the variational RNN (VRNN)1 can model the kind of variability observed in highly structured sequential data such as natural speech. We empirically evaluate the proposed model against related sequential models on four speech datasets and one handwriting dataset. Our results show the important roles that latent random variables can play in the RNN dynamic hidden state. | A Recurrent Latent Variable Model for Sequential Data | 2,911 |
We first present a general risk bound for ensembles that depends on the Lp norm of the weighted combination of voters which can be selected from a continuous set. We then propose a boosting method, called QuadBoost, which is strongly supported by the general risk bound and has very simple rules for assigning the voters' weights. Moreover, QuadBoost exhibits a rate of decrease of its empirical error which is slightly faster than the one achieved by AdaBoost. The experimental results confirm the expectation of the theory that QuadBoost is a very efficient method for learning ensembles. | Efficient Learning of Ensembles with QuadBoost | 2,912 |
We consider emphatic temporal-difference learning algorithms for policy evaluation in discounted Markov decision processes with finite spaces. Such algorithms were recently proposed by Sutton, Mahmood, and White (2015) as an improved solution to the problem of divergence of off-policy temporal-difference learning with linear function approximation. We present in this paper the first convergence proofs for two emphatic algorithms, ETD($\lambda$) and ELSTD($\lambda$). We prove, under general off-policy conditions, the convergence in $L^1$ for ELSTD($\lambda$) iterates, and the almost sure convergence of the approximate value functions calculated by both algorithms using a single infinitely long trajectory. Our analysis involves new techniques with applications beyond emphatic algorithms leading, for example, to the first proof that standard TD($\lambda$) also converges under off-policy training for $\lambda$ sufficiently large. | On Convergence of Emphatic Temporal-Difference Learning | 2,913 |
In this paper, a novel framework of sparse kernel learning for Support Vector Data Description (SVDD) based anomaly detection is presented. In this work, optimal sparse feature selection for anomaly detection is first modeled as a Mixed Integer Programming (MIP) problem. Due to the prohibitively high computational complexity of the MIP, it is relaxed into a Quadratically Constrained Linear Programming (QCLP) problem. The QCLP problem can then be practically solved by using an iterative optimization method, in which multiple subsets of features are iteratively found as opposed to a single subset. The QCLP-based iterative optimization problem is solved in a finite space called the \emph{Empirical Kernel Feature Space} (EKFS) instead of in the input space or \emph{Reproducing Kernel Hilbert Space} (RKHS). This is possible because of the fact that the geometrical properties of the EKFS and the corresponding RKHS remain the same. Now, an explicit nonlinear exploitation of the data in a finite EKFS is achievable, which results in optimal feature ranking. Experimental results based on a hyperspectral image show that the proposed method can provide improved performance over the current state-of-the-art techniques. | Optimal Sparse Kernel Learning for Hyperspectral Anomaly Detection | 2,914 |
We study the interpretability of conditional probability estimates for binary classification under the agnostic setting or scenario. Under the agnostic setting, conditional probability estimates do not necessarily reflect the true conditional probabilities. Instead, they have a certain calibration property: among all data points that the classifier has predicted P(Y = 1|X) = p, p portion of them actually have label Y = 1. For cost-sensitive decision problems, this calibration property provides adequate support for us to use Bayes Decision Theory. In this paper, we define a novel measure for the calibration property together with its empirical counterpart, and prove an uniform convergence result between them. This new measure enables us to formally justify the calibration property of conditional probability estimations, and provides new insights on the problem of estimating and calibrating conditional probabilities. | On the Interpretability of Conditional Probability Estimates in the
Agnostic Setting | 2,915 |
The outputs of non-linear feed-forward neural network are positive, which could be treated as probability when they are normalized to one. If we take Entropy-Based Principle into consideration, the outputs for each sample could be represented as the distribution of this sample for different clusters. Entropy-Based Principle is the principle with which we could estimate the unknown distribution under some limited conditions. As this paper defines two processes in Feed-Forward Neural Network, our limited condition is the abstracted features of samples which are worked out in the abstraction process. And the final outputs are the probability distribution for different clusters in the clustering process. As Entropy-Based Principle is considered into the feed-forward neural network, a clustering method is born. We have conducted some experiments on six open UCI datasets, comparing with a few baselines and applied purity as the measurement . The results illustrate that our method outperforms all the other baselines that are most popular clustering methods. | Max-Entropy Feed-Forward Clustering Neural Network | 2,916 |
Margin-Based Principle has been proposed for a long time, it has been proved that this principle could reduce the structural risk and improve the performance in both theoretical and practical aspects. Meanwhile, feed-forward neural network is a traditional classifier, which is very hot at present with a deeper architecture. However, the training algorithm of feed-forward neural network is developed and generated from Widrow-Hoff Principle that means to minimize the squared error. In this paper, we propose a new training algorithm for feed-forward neural networks based on Margin-Based Principle, which could effectively promote the accuracy and generalization ability of neural network classifiers with less labelled samples and flexible network. We have conducted experiments on four UCI open datasets and achieved good results as expected. In conclusion, our model could handle more sparse labelled and more high-dimension dataset in a high accuracy while modification from old ANN method to our method is easy and almost free of work. | Margin-Based Feed-Forward Neural Network Classifiers | 2,917 |
In this document, we show that the algorithm CoCoA+ (Ma et al., ICML, 2015) under the setting used in their experiments, which is also the best setting suggested by the authors that proposed this algorithm, is equivalent to the practical variant of DisDCA (Yang, NIPS, 2013). | On the Equivalence of CoCoA+ and DisDCA | 2,918 |
This paper studies the generalization performance of multi-class classification algorithms, for which we obtain, for the first time, a data-dependent generalization error bound with a logarithmic dependence on the class size, substantially improving the state-of-the-art linear dependence in the existing data-dependent generalization analysis. The theoretical analysis motivates us to introduce a new multi-class classification machine based on $\ell_p$-norm regularization, where the parameter $p$ controls the complexity of the corresponding bounds. We derive an efficient optimization algorithm based on Fenchel duality theory. Benchmarks on several real-world datasets show that the proposed algorithm can achieve significant accuracy gains over the state of the art. | Multi-class SVMs: From Tighter Data-Dependent Generalization Bounds to
Novel Algorithms | 2,919 |
We propose a localized approach to multiple kernel learning that can be formulated as a convex optimization problem over a given cluster structure. For which we obtain generalization error guarantees and derive an optimization algorithm based on the Fenchel dual representation. Experiments on real-world datasets from the application domains of computational biology and computer vision show that convex localized multiple kernel learning can achieve higher prediction accuracies than its global and non-convex local counterparts. | Localized Multiple Kernel Learning---A Convex Approach | 2,920 |
This work builds upon previous efforts in online incremental learning, namely the Incremental Gaussian Mixture Network (IGMN). The IGMN is capable of learning from data streams in a single-pass by improving its model after analyzing each data point and discarding it thereafter. Nevertheless, it suffers from the scalability point-of-view, due to its asymptotic time complexity of $\operatorname{O}\bigl(NKD^3\bigr)$ for $N$ data points, $K$ Gaussian components and $D$ dimensions, rendering it inadequate for high-dimensional data. In this paper, we manage to reduce this complexity to $\operatorname{O}\bigl(NKD^2\bigr)$ by deriving formulas for working directly with precision matrices instead of covariance matrices. The final result is a much faster and scalable algorithm which can be applied to high dimensional tasks. This is confirmed by applying the modified algorithm to high-dimensional classification datasets. | A Fast Incremental Gaussian Mixture Model | 2,921 |
The online learning of deep neural networks is an interesting problem of machine learning because, for example, major IT companies want to manage the information of the massive data uploaded on the web daily, and this technology can contribute to the next generation of lifelong learning. We aim to train deep models from new data that consists of new classes, distributions, and tasks at minimal computational cost, which we call online deep learning. Unfortunately, deep neural network learning through classical online and incremental methods does not work well in both theory and practice. In this paper, we introduce dual memory architectures for online incremental deep learning. The proposed architecture consists of deep representation learners and fast learnable shallow kernel networks, both of which synergize to track the information of new data. During the training phase, we use various online, incremental ensemble, and transfer learning techniques in order to achieve lower error of the architecture. On the MNIST, CIFAR-10, and ImageNet image recognition tasks, the proposed dual memory architectures performs much better than the classical online and incremental ensemble algorithm, and their accuracies are similar to that of the batch learner. | Dual Memory Architectures for Fast Deep Learning of Stream Data via an
Online-Incremental-Transfer Strategy | 2,922 |
We present doubly stochastic gradient MCMC, a simple and generic method for (approximate) Bayesian inference of deep generative models (DGMs) in a collapsed continuous parameter space. At each MCMC sampling step, the algorithm randomly draws a mini-batch of data samples to estimate the gradient of log-posterior and further estimates the intractable expectation over hidden variables via a neural adaptive importance sampler, where the proposal distribution is parameterized by a deep neural network and learnt jointly. We demonstrate the effectiveness on learning various DGMs in a wide range of tasks, including density estimation, data generation and missing data imputation. Our method outperforms many state-of-the-art competitors. | Learning Deep Generative Models with Doubly Stochastic MCMC | 2,923 |
Deep directed generative models have attracted much attention recently due to their expressive representation power and the ability of ancestral sampling. One major difficulty of learning directed models with many latent variables is the intractable inference. To address this problem, most existing algorithms make assumptions to render the latent variables independent of each other, either by designing specific priors, or by approximating the true posterior using a factorized distribution. We believe the correlations among latent variables are crucial for faithful data representation. Driven by this idea, we propose an inference method based on the conditional pseudo-likelihood that preserves the dependencies among the latent variables. For learning, we propose to employ the hard Expectation Maximization (EM) algorithm, which avoids the intractability of the traditional EM by max-out instead of sum-out to compute the data likelihood. Qualitative and quantitative evaluations of our model against state of the art deep models on benchmark datasets demonstrate the effectiveness of the proposed algorithm in data representation and reconstruction. | Latent Regression Bayesian Network for Data Representation | 2,924 |
We consider stochastic sequential learning problems where the learner can observe the \textit{average reward of several actions}. Such a setting is interesting in many applications involving monitoring and surveillance, where the set of the actions to observe represent some (geographical) area. The importance of this setting is that in these applications, it is actually \textit{cheaper} to observe average reward of a group of actions rather than the reward of a single action. We show that when the reward is \textit{smooth} over a given graph representing the neighboring actions, we can maximize the cumulative reward of learning while \textit{minimizing the sensing cost}. In this paper we propose CheapUCB, an algorithm that matches the regret guarantees of the known algorithms for this setting and at the same time guarantees a linear cost again over them. As a by-product of our analysis, we establish a $\Omega(\sqrt{dT})$ lower bound on the cumulative regret of spectral bandits for a class of graphs with effective dimension $d$. | Cheap Bandits | 2,925 |
We extend the theory of boosting for regression problems to the online learning setting. Generalizing from the batch setting for boosting, the notion of a weak learning algorithm is modeled as an online learning algorithm with linear loss functions that competes with a base class of regression functions, while a strong learning algorithm is an online learning algorithm with convex loss functions that competes with a larger class of regression functions. Our main result is an online gradient boosting algorithm which converts a weak online learning algorithm into a strong one where the larger class of functions is the linear span of the base class. We also give a simpler boosting algorithm that converts a weak online learning algorithm into a strong one where the larger class of functions is the convex hull of the base class, and prove its optimality. | Online Gradient Boosting | 2,926 |
We study supervised learning problems using clustering constraints to impose structure on either features or samples, seeking to help both prediction and interpretation. The problem of clustering features arises naturally in text classification for instance, to reduce dimensionality by grouping words together and identify synonyms. The sample clustering problem on the other hand, applies to multiclass problems where we are allowed to make multiple predictions and the performance of the best answer is recorded. We derive a unified optimization formulation highlighting the common structure of these problems and produce algorithms whose core iteration complexity amounts to a k-means clustering step, which can be approximated efficiently. We extend these results to combine sparsity and clustering constraints, and develop a new projection algorithm on the set of clustered sparse vectors. We prove convergence of our algorithms on random instances, based on a union of subspaces interpretation of the clustering structure. Finally, we test the robustness of our methods on artificial data sets as well as real data extracted from movie reviews. | Learning with Clustering Structure | 2,927 |
Most work in the area of statistical relational learning (SRL) is focussed on discrete data, even though a few approaches for hybrid SRL models have been proposed that combine numerical and discrete variables. In this paper we distinguish numerical random variables for which a probability distribution is defined by the model from numerical input variables that are only used for conditioning the distribution of discrete response variables. We show how numerical input relations can very easily be used in the Relational Bayesian Network framework, and that existing inference and learning methods need only minor adjustments to be applied in this generalized setting. The resulting framework provides natural relational extensions of classical probabilistic models for categorical data. We demonstrate the usefulness of RBN models with numeric input relations by several examples. In particular, we use the augmented RBN framework to define probabilistic models for multi-relational (social) networks in which the probability of a link between two nodes depends on numeric latent feature vectors associated with the nodes. A generic learning procedure can be used to obtain a maximum-likelihood fit of model parameters and latent feature values for a variety of models that can be expressed in the high-level RBN representation. Specifically, we propose a model that allows us to interpret learned latent feature values as community centrality degrees by which we can identify nodes that are central for one community, that are hubs between communities, or that are isolated nodes. In a multi-relational setting, the model also provides a characterization of how different relations are associated with each community. | Numeric Input Relations for Relational Learning with Applications to
Community Structure Analysis | 2,928 |
People believe that depth plays an important role in success of deep neural networks (DNN). However, this belief lacks solid theoretical justifications as far as we know. We investigate role of depth from perspective of margin bound. In margin bound, expected error is upper bounded by empirical margin error plus Rademacher Average (RA) based capacity term. First, we derive an upper bound for RA of DNN, and show that it increases with increasing depth. This indicates negative impact of depth on test performance. Second, we show that deeper networks tend to have larger representation power (measured by Betti numbers based complexity) than shallower networks in multi-class setting, and thus can lead to smaller empirical margin error. This implies positive impact of depth. The combination of these two results shows that for DNN with restricted number of hidden units, increasing depth is not always good since there is a tradeoff between positive and negative impacts. These results inspire us to seek alternative ways to achieve positive impact of depth, e.g., imposing margin-based penalty terms to cross entropy loss so as to reduce empirical margin error without increasing depth. Our experiments show that in this way, we achieve significantly better test performance. | On the Depth of Deep Neural Networks: A Theoretical View | 2,929 |
In a variety of problems originating in supervised, unsupervised, and reinforcement learning, the loss function is defined by an expectation over a collection of random variables, which might be part of a probabilistic model or the external world. Estimating the gradient of this loss function, using samples, lies at the core of gradient-based learning algorithms for these problems. We introduce the formalism of stochastic computation graphs---directed acyclic graphs that include both deterministic functions and conditional probability distributions---and describe how to easily and automatically derive an unbiased estimator of the loss function's gradient. The resulting algorithm for computing the gradient estimator is a simple modification of the standard backpropagation algorithm. The generic scheme we propose unifies estimators derived in variety of prior work, along with variance-reduction techniques therein. It could assist researchers in developing intricate models involving a combination of stochastic and deterministic operations, enabling, for example, attention, memory, and control actions. | Gradient Estimation Using Stochastic Computation Graphs | 2,930 |
We present and empirically evaluate an efficient algorithm that learns to aggregate the predictions of an ensemble of binary classifiers. The algorithm uses the structure of the ensemble predictions on unlabeled data to yield significant performance improvements. It does this without making assumptions on the structure or origin of the ensemble, without parameters, and as scalably as linear learning. We empirically demonstrate these performance gains with random forests. | Scalable Semi-Supervised Aggregation of Classifiers | 2,931 |
It is often desirable to be able to recognize when inputs to a recognition function learned in a supervised manner correspond to classes unseen at training time. With this ability, new class labels could be assigned to these inputs by a human operator, allowing them to be incorporated into the recognition function --- ideally under an efficient incremental update mechanism. While good algorithms that assume inputs from a fixed set of classes exist, e.g., artificial neural networks and kernel machines, it is not immediately obvious how to extend them to perform incremental learning in the presence of unknown query classes. Existing algorithms take little to no distributional information into account when learning recognition functions and lack a strong theoretical foundation. We address this gap by formulating a novel, theoretically sound classifier --- the Extreme Value Machine (EVM). The EVM has a well-grounded interpretation derived from statistical Extreme Value Theory (EVT), and is the first classifier to be able to perform nonlinear kernel-free variable bandwidth incremental learning. Compared to other classifiers in the same deep network derived feature space, the EVM is accurate and efficient on an established benchmark partition of the ImageNet dataset. | The Extreme Value Machine | 2,932 |
Machine learning relies on the assumption that unseen test instances of a classification problem follow the same distribution as observed training data. However, this principle can break down when machine learning is used to make important decisions about the welfare (employment, education, health) of strategic individuals. Knowing information about the classifier, such individuals may manipulate their attributes in order to obtain a better classification outcome. As a result of this behavior---often referred to as gaming---the performance of the classifier may deteriorate sharply. Indeed, gaming is a well-known obstacle for using machine learning methods in practice; in financial policy-making, the problem is widely known as Goodhart's law. In this paper, we formalize the problem, and pursue algorithms for learning classifiers that are robust to gaming. We model classification as a sequential game between a player named "Jury" and a player named "Contestant." Jury designs a classifier, and Contestant receives an input to the classifier, which he may change at some cost. Jury's goal is to achieve high classification accuracy with respect to Contestant's original input and some underlying target classification function. Contestant's goal is to achieve a favorable classification outcome while taking into account the cost of achieving it. For a natural class of cost functions, we obtain computationally efficient learning algorithms which are near-optimal. Surprisingly, our algorithms are efficient even on concept classes that are computationally hard to learn. For general cost functions, designing an approximately optimal strategy-proof classifier, for inverse-polynomial approximation, is NP-hard. | Strategic Classification | 2,933 |
We tackle the problem of learning linear classifiers from noisy datasets in a multiclass setting. The two-class version of this problem was studied a few years ago where the proposed approaches to combat the noise revolve around a Per-ceptron learning scheme fed with peculiar examples computed through a weighted average of points from the noisy training set. We propose to build upon these approaches and we introduce a new algorithm called UMA (for Unconfused Multiclass additive Algorithm) which may be seen as a generalization to the multiclass setting of the previous approaches. In order to characterize the noise we use the confusion matrix as a multiclass extension of the classification noise studied in the aforemen-tioned literature. Theoretically well-founded, UMA furthermore displays very good empirical noise robustness, as evidenced by numerical simulations conducted on both synthetic and real data. | Unconfused ultraconservative multiclass algorithms | 2,934 |
The computational cost of many signal processing and machine learning techniques is often dominated by the cost of applying certain linear operators to high-dimensional vectors. This paper introduces an algorithm aimed at reducing the complexity of applying linear operators in high dimension by approximately factorizing the corresponding matrix into few sparse factors. The approach relies on recent advances in non-convex optimization. It is first explained and analyzed in details and then demonstrated experimentally on various problems including dictionary learning for image denoising, and the approximation of large matrices arising in inverse problems. | Flexible Multi-layer Sparse Approximations of Matrices and Applications | 2,935 |
Stochastic algorithms are efficient approaches to solving machine learning and optimization problems. In this paper, we propose a general framework called Splash for parallelizing stochastic algorithms on multi-node distributed systems. Splash consists of a programming interface and an execution engine. Using the programming interface, the user develops sequential stochastic algorithms without concerning any detail about distributed computing. The algorithm is then automatically parallelized by a communication-efficient execution engine. We provide theoretical justifications on the optimal rate of convergence for parallelizing stochastic gradient descent. Splash is built on top of Apache Spark. The real-data experiments on logistic regression, collaborative filtering and topic modeling verify that Splash yields order-of-magnitude speedup over single-thread stochastic algorithms and over state-of-the-art implementations on Spark. | Splash: User-friendly Programming Interface for Parallelizing Stochastic
Algorithms | 2,936 |
We discuss necessary and sufficient conditions for an auto-encoder to define a conservative vector field, in which case it is associated with an energy function akin to the unnormalized log-probability of the data. We show that the conditions for conservativeness are more general than for encoder and decoder weights to be the same ("tied weights"), and that they also depend on the form of the hidden unit activation function, but that contractive training criteria, such as denoising, will enforce these conditions locally. Based on these observations, we show how we can use auto-encoders to extract the conservative component of a vector field. | Conservativeness of untied auto-encoders | 2,937 |
We present a complimentary objective for training recurrent neural networks (RNN) with gating units that helps with regularization and interpretability of the trained model. Attention-based RNN models have shown success in many difficult sequence to sequence classification problems with long and short term dependencies, however these models are prone to overfitting. In this paper, we describe how to regularize these models through an L1 penalty on the activation of the gating units, and show that this technique reduces overfitting on a variety of tasks while also providing to us a human-interpretable visualization of the inputs used by the network. These tasks include sentiment analysis, paraphrase recognition, and question answering. | Occam's Gates | 2,938 |
This paper presents a multiscale decomposition algorithm. Unlike standard wavelet transforms, the proposed operator is both linear and shift invariant. The central idea is to obtain shift invariance by averaging the aligned wavelet transform projections over all circular shifts of the signal. It is shown how the same transform can be obtained by a linear filter bank. | A Linear Shift Invariant Multiscale Transform | 2,939 |
We give a systematic, abstract formulation of the image normalization method as applied to a general group of image transformations, and then illustrate the abstract analysis by applying it to the hierarchy of viewing transformations of a planar object. | General Theory of Image Normalization | 2,940 |
This paper presents an invariant under scaling and linear brightness change. The invariant is based on differentials and therefore is a local feature. Rotationally invariant 2-d differential Gaussian operators up to third order are proposed for the implementation of the invariant. The performance is analyzed by simulating a camera zoom-out. | A Differential Invariant for Zooming | 2,941 |
We describe a simple, but efficient algorithm for the generation of dilated contours from bilevel images. The initial part of the contour extraction is explained to be a good candidate for parallel computer code generation. The remainder of the algorithm is of linear nature. | A Parallel Algorithm for Dilated Contour Extraction from Bilevel Images | 2,942 |
Fractal image compression, Culik's image compression and zerotree prediction coding of wavelet image decomposition coefficients succeed only because typical images being compressed possess a significant degree of self-similarity. Besides the common concept, these methods turn out to be even more tightly related, to the point of algorithmical reducibility of one technique to another. The goal of the present paper is to demonstrate these relations. The paper offers a plain-term interpretation of Culik's image compression, in regular image processing terms, without resorting to finite state machines and similar lofty language. The interpretation is shown to be algorithmically related to an IFS fractal image compression method: an IFS can be exactly transformed into Culik's image code. Using this transformation, we will prove that in a self-similar (part of an) image any zero wavelet coefficient is the root of a zerotree, or its branch. The paper discusses the zerotree coding of (wavelet/projection) coefficients as a common predictor/corrector, applied vertically through different layers of a multiresolutional decomposition, rather than within the same view. This interpretation leads to an insight into the evolution of image compression techniques: from a causal single-layer prediction, to non-causal same-view predictions (wavelet decomposition among others) and to a causal cross-layer prediction (zero-trees, Culik's method). | Image Compression with Iterated Function Systems, Finite Automata and
Zerotrees: Grand Unification | 2,943 |
This paper presents invariants under gamma correction and similarity transformations. The invariants are local features based on differentials which are implemented using derivatives of the Gaussian. The use of the proposed invariant representation is shown to yield improved correlation results in a template matching scenario. | Differential Invariants under Gamma Correction | 2,944 |
This work deals with content-based video indexing. Our viewpoint is semi-automatic analysis of compressed video. We consider the possible applications of motion analysis and moving object detection : assisting moving object indexing, summarising videos, and allowing image and motion queries. We propose an approach based on interest points. As first results, we test and compare the stability of different types of interest point detectors in compressed sequences. | Assisted Video Sequences Indexing : Motion Analysis Based on Interest
Points | 2,945 |
The paper has established and verified the theory prevailing widely among image and pattern recognition specialists that the bottom-up indirect regional matching process is the more stable and the more robust than the global matching process against concentrated types of noise represented by clutter, outlier or occlusion in the imagery. We have demonstrated this by analyzing the effect of concentrated noise on a typical decision making process of a simplified two candidate voting model where our theorem establishes the lower bounds to a critical breakdown point of election (or decision) result by the bottom-up matching process are greater than the exact bound of the global matching process implying that the former regional process is capable of accommodating a higher level of noise than the latter global process before the result of decision overturns. We present a convincing experimental verification supporting not only the theory by a white-black flag recognition problem in the presence of localized noise but also the validity of the conjecture by a facial recognition problem that the theorem remains valid for other decision making processes involving an important dimension-reducing transform such as principal component analysis or a Gabor transform. | Robustness of Regional Matching Scheme over Global Matching Scheme | 2,946 |
A Bayesian classifier that up-weights the differences in the attribute values is discussed. Using four popular datasets from the UCI repository, some interesting features of the network are illustrated. The network is suitable for classification problems. | Boosting the Differences: A fast Bayesian classifier neural network | 2,947 |
The difference-boosting algorithm is used on letters dataset from the UCI repository to classify distorted raster images of English alphabets. In contrast to rather complex networks, the difference-boosting is found to produce comparable or better classification efficiency on this complex problem. | Distorted English Alphabet Identification : An application of Difference
Boosting Algorithm | 2,948 |
We present a new method to transform the spectral pixel information of a micrograph into an affine geometric description, which allows us to analyze the morphology of granular materials. We use spectral and pulse-coupled neural network based segmentation techniques to generate blobs, and a newly developed algorithm to extract dilated contours. A constrained Delaunay tesselation of the contour points results in a triangular mesh. This mesh is the basic ingredient of the Chodal Axis Transform, which provides a morphological decomposition of shapes. Such decomposition allows for grain separation and the efficient computation of the statistical features of granular materials. | Geometric Morphology of Granular Materials | 2,949 |
The problem of searching for a model-based scene interpretation is analyzed within a probabilistic framework. Object models are formulated as generative models for range data of the scene. A new statistical criterion, the truncated object probability, is introduced to infer an optimal sequence of object hypotheses to be evaluated for their match to the data. The truncated probability is partly determined by prior knowledge of the objects and partly learned from data. Some experiments on sequence quality and object segmentation and recognition from stereo data are presented. The article recovers classic concepts from object recognition (grouping, geometric hashing, alignment) from the probabilistic perspective and adds insight into the optimal ordering of object hypotheses for evaluation. Moreover, it introduces point-relation densities, a key component of the truncated probability, as statistical models of local surface shape. | Probabilistic Search for Object Segmentation and Recognition | 2,950 |
We study theoretical and computational aspects of the least squares fit (LSF) of circles and circular arcs. First we discuss the existence and uniqueness of LSF and various parametrization schemes. Then we evaluate several popular circle fitting algorithms and propose a new one that surpasses the existing methods in reliability. We also discuss and compare direct (algebraic) circle fits. | Least squares fitting of circles and lines | 2,951 |
We study the problem of fitting parametrized curves to noisy data. Under certain assumptions (known as Cartesian and radial functional models), we derive asymptotic expressions for the bias and the covariance matrix of the parameter estimates. We also extend Kanatani's version of the Cramer-Rao lower bound, which he proved for unbiased estimates only, to more general estimates that include many popular algorithms (most notably, the orthogonal least squares and algebraic fits). We then show that the gradient-weighted algebraic fit is statistically efficient and describe all other statistically efficient algebraic fits. | Statistical efficiency of curve fitting algorithms | 2,952 |
Most algorithms in 3D computer vision rely on the pinhole camera model because of its simplicity, whereas virtually all imaging devices introduce certain amount of nonlinear distortion, where the radial distortion is the most severe part. Common approach to radial distortion is by the means of polynomial approximation, which introduces distortion-specific parameters into the camera model and requires estimation of these distortion parameters. The task of estimating radial distortion is to find a radial distortion model that allows easy undistortion as well as satisfactory accuracy. This paper presents a new radial distortion model with an easy analytical undistortion formula, which also belongs to the polynomial approximation category. Experimental results are presented to show that with this radial distortion model, satisfactory accuracy is achieved. An application of the new radial distortion model is non-iterative yellow line alignment with a calibrated camera on ODIS, a robot built in our CSOIS. | Flexible Camera Calibration Using a New Analytical Radial Undistortion
Formula with Application to Mobile Robot Localization | 2,953 |
Common approach to radial distortion is by the means of polynomial approximation, which introduces distortion-specific parameters into the camera model and requires estimation of these distortion parameters. The task of estimating radial distortion is to find a radial distortion model that allows easy undistortion as well as satisfactory accuracy. This paper presents a new radial distortion model with an easy analytical undistortion formula, which also belongs to the polynomial approximation category. Experimental results are presented to show that with this radial distortion model, satisfactory accuracy is achieved. | A New Analytical Radial Distortion Model for Camera Calibration | 2,954 |
The common approach to radial distortion is by the means of polynomial approximation, which introduces distortion-specific parameters into the camera model and requires estimation of these distortion parameters. The task of estimating radial distortion is to find a radial distortion model that allows easy undistortion as well as satisfactory accuracy. This paper presents a new class of rational radial distortion models with easy analytical undistortion formulae. Experimental results are presented to show that with this class of rational radial distortion models, satisfactory and comparable accuracy is achieved. | Rational Radial Distortion Models with Analytical Undistortion Formulae | 2,955 |
The common approach to radial distortion is by the means of polynomial approximation, which introduces distortion-specific parameters into the camera model and requires estimation of these distortion parameters. The task of estimating radial distortion is to find a radial distortion model that allows easy undistortion as well as satisfactory accuracy. This paper presents a new piecewise radial distortion model with easy analytical undistortion formula. The motivation for seeking a piecewise radial distortion model is that, when a camera is resulted in a low quality during manufacturing, the nonlinear radial distortion can be complex. Using low order polynomials to approximate the radial distortion might not be precise enough. On the other hand, higher order polynomials suffer from the inverse problem. With the new piecewise radial distortion function, more flexibility is obtained and the radial undistortion can be performed analytically. Experimental results are presented to show that with this new piecewise radial distortion model, better performance is achieved than that using the single function. Furthermore, a comparable performance with the conventional polynomial model using 2 coefficients can also be accomplished. | An Analytical Piecewise Radial Distortion Model for Precision Camera
Calibration | 2,956 |
The task of camera calibration is to estimate the intrinsic and extrinsic parameters of a camera model. Though there are some restricted techniques to infer the 3-D information about the scene from uncalibrated cameras, effective camera calibration procedures will open up the possibility of using a wide range of existing algorithms for 3-D reconstruction and recognition. The applications of camera calibration include vision-based metrology, robust visual platooning and visual docking of mobile robots where the depth information is important. | Camera Calibration: a USU Implementation | 2,957 |
The commonly used radial distortion model for camera calibration is in fact an assumption or a restriction. In practice, camera distortion could happen in a general geometrical manner that is not limited to the radial sense. This paper proposes a simplified geometrical distortion modeling method by using two different radial distortion functions in the two image axes. A family of simplified geometric distortion models is proposed, which are either simple polynomials or the rational functions of polynomials. Analytical geometric undistortion is possible using two of the distortion functions discussed in this paper and their performance can be improved by applying a piecewise fitting idea. Our experimental results show that the geometrical distortion models always perform better than their radial distortion counterparts. Furthermore, the proposed geometric modeling method is more appropriate for cameras whose distortion is not perfectly radially symmetric around the center of distortion. | A Family of Simplified Geometric Distortion Models for Camera
Calibration | 2,958 |
In the paper will be presented a safety and security system based on fingerprint technology. The results suggest a new scenario where the new cars can use a fingerprint sensor integrated in car handle to allow access and in the dashboard as starter button. | Fingerprint based bio-starter and bio-access | 2,959 |
For many tracking and surveillance applications, background subtraction provides an effective means of segmenting objects moving in front of a static background. Researchers have traditionally used combinations of morphological operations to remove the noise inherent in the background-subtracted result. Such techniques can effectively isolate foreground objects, but tend to lose fidelity around the borders of the segmentation, especially for noisy input. This paper explores the use of a minimum graph cut algorithm to segment the foreground, resulting in qualitatively and quantitiatively cleaner segmentations. Experiments on both artificial and real data show that the graph-based method reduces the error around segmented foreground objects. A MATLAB code implementation is available at http://www.cs.smith.edu/~nhowe/research/code/#fgseg | Better Foreground Segmentation Through Graph Cuts | 2,960 |
A method of temporal factor prognosis of TE (tick-borne encephalitis) infection has been developed. The high precision of the prognosis results for a number of geographical regions of Primorsky Krai has been achieved. The method can be applied not only to epidemiological research but also to others. | Factor Temporal Prognosis of Tick-Borne Encephalitis Foci Functioning on
the South of Russian Far East | 2,961 |
Despite encouraging recent progresses in ensemble approaches, classification methods seem to have reached a plateau in development. Further advances depend on a better understanding of geometrical and topological characteristics of point sets in high-dimensional spaces, the preservation of such characteristics under feature transformations and sampling processes, and their interaction with geometrical models used in classifiers. We discuss an attempt to measure such properties from data sets and relate them to classifier accuracies. | Geometrical Complexity of Classification Problems | 2,962 |
This publication presents methods for face detection, analysis and recognition: fast normalized cross-correlation (fast correlation coefficient) between multiple templates based face pre-detection method, method for detection of exact face contour based on snakes and Generalized Gradient Vector Flow field, method for combining recognition algorithms based on Cumulative Match Characteristics in order to increase recognition speed and accuracy, and face recognition method based on Principal Component Analysis of the Wavelet Packet Decomposition allowing to use PCA - based recognition method with large number of training images. For all the methods are presented experimental results and comparisons of speed and accuracy with large face databases. | Computerized Face Detection and Recognition | 2,963 |
This paper presents a blind detection and compensation technique for camera lens geometric distortions. The lens distortion introduces higher-order correlations in the frequency domain and in turn it can be detected using higher-order spectral analysis tools without assuming any specific calibration target. The existing blind lens distortion removal method only considered a single-coefficient radial distortion model. In this paper, two coefficients are considered to model approximately the geometric distortion. All the models considered have analytical closed-form inverse formulae. | Blind Detection and Compensation of Camera Lens Geometric Distortions | 2,964 |
We study image compression by a separable wavelet basis $\big\{\psi(2^{k_1}x-i)\psi(2^{k_2}y-j),$ $\phi(x-i)\psi(2^{k_2}y-j),$ $\psi(2^{k_1}(x-i)\phi(y-j),$ $\phi(x-i)\phi(y-i)\big\},$ where $k_1, k_2 \in \mathbb{Z}_+$; $i,j\in\mathbb{Z}$; and $\phi,\psi$ are elements of a standard biorthogonal wavelet basis in $L_2(\mathbb{R})$. Because $k_1\ne k_2$, the supports of the basis elements are rectangles, and the corresponding transform is known as the {\em rectangular wavelet transform}. We prove that if one-dimensional wavelet basis has $M$ dual vanishing moments then the rate of approximation by $N$ coefficients of rectangular wavelet transform is $\mathcal{O}(N^{-M}\log^C N)$ for functions with mixed derivative of order $M$ in each direction. The square wavelet transform yields the approximation rate is $\mathcal{O}(N^{-M/2})$ for functions with all derivatives of the total order $M$. Thus, the rectangular wavelet transform can outperform the square one if an image has a mixed derivative. We provide experimental comparison of image compression which shows that rectangular wavelet transform outperform the square one. | Image compression by rectangular wavelet transform | 2,965 |
The Gradient Vector Flow (GVF) is a vector diffusion approach based on Partial Differential Equations (PDEs). This method has been applied together with snake models for boundary extraction medical images segmentation. The key idea is to use a diffusion-reaction PDE to generate a new external force field that makes snake models less sensitivity to initialization as well as improves the snake's ability to move into boundary concavities. In this paper, we firstly review basic results about convergence and numerical analysis of usual GVF schemes. We point out that GVF presents numerical problems due to discontinuities image intensity. This point is considered from a practical viewpoint from which the GVF parameters must follow a relationship in order to improve numerical convergence. Besides, we present an analytical analysis of the GVF dependency from the parameters values. Also, we observe that the method can be used for multiply connected domains by just imposing the suitable boundary condition. In the experimental results we verify these theoretical points and demonstrate the utility of GVF on a segmentation approach that we have developed based on snakes. | Gradient Vector Flow Models for Boundary Extraction in 2D Images | 2,966 |
Image information content is known to be a complicated and controvercial problem. This paper posits a new image information content definition. Following the theory of Solomonoff-Kolmogorov-Chaitin's complexity, we define image information content as a set of descriptions of imafe data structures. Three levels of such description can be generally distinguished: 1)the global level, where the coarse structure of the entire scene is initially outlined; 2) the intermediate level, where structures of separate, non-overlapping image regions usually associated with individual scene objects are deliniated; and 3) the low-level description, where local image structures observed in a limited and restricted field of view are resolved. A technique for creating such image information content descriptors is developed. Its algorithm is presented and elucidated with some examples, which demonstrate the effectiveness of the proposed approach. | Searching for image information content, its discovery, extraction, and
representation | 2,967 |
In this paper we present an unconventional image segmentation approach which is devised to meet the requirements of image understanding and pattern recognition tasks. Generally image understanding assumes interplay of two sub-processes: image information content discovery and image information content interpretation. Despite of its widespread use, the notion of "image information content" is still ill defined, intuitive, and ambiguous. Most often, it is used in the Shannon's sense, which means information content assessment averaged over the whole signal ensemble. Humans, however,rarely resort to such estimates. They are very effective in decomposing images into their meaningful constituents and focusing attention to the perceptually relevant image parts. We posit that following the latest findings in human attention vision studies and the concepts of Kolmogorov's complexity theory an unorthodox segmentation approach can be proposed that provides effective image decomposition to information preserving image fragments well suited for subsequent image interpretation. We provide some illustrative examples, demonstrating effectiveness of this approach. | Paving the Way for Image Understanding: A New Kind of Image
Decomposition is Desired | 2,968 |
We present an automatic face verification system inspired by known properties of biological systems. In the proposed algorithm the whole image is converted from the spatial to polar frequency domain by a Fourier-Bessel Transform (FBT). Using the whole image is compared to the case where only face image regions (local analysis) are considered. The resulting representations are embedded in a dissimilarity space, where each image is represented by its distance to all the other images, and a Pseudo-Fisher discriminator is built. Verification test results on the FERET database showed that the local-based algorithm outperforms the global-FBT version. The local-FBT algorithm performed as state-of-the-art methods under different testing conditions, indicating that the proposed system is highly robust for expression, age, and illumination variations. We also evaluated the performance of the proposed system under strong occlusion conditions and found that it is highly robust for up to 50% of face occlusion. Finally, we automated completely the verification system by implementing face and eye detection algorithms. Under this condition, the local approach was only slightly superior to the global approach. | Automatic Face Recognition System Based on Local Fourier-Bessel Features | 2,969 |
A novel biologically motivated face recognition algorithm based on polar frequency is presented. Polar frequency descriptors are extracted from face images by Fourier-Bessel transform (FBT). Next, the Euclidean distance between all images is computed and each image is now represented by its dissimilarity to the other images. A Pseudo-Fisher Linear Discriminant was built on this dissimilarity space. The performance of Discrete Fourier transform (DFT) descriptors, and a combination of both feature types was also evaluated. The algorithms were tested on a 40- and 1196-subjects face database (ORL and FERET, respectively). With 5 images per subject in the training and test datasets, error rate on the ORL database was 3.8, 1.25 and 0.2% for the FBT, DFT, and the combined classifier, respectively, as compared to 2.6% achieved by the best previous algorithm. The most informative polar frequency features were concentrated at low-to-medium angular frequencies coupled to low radial frequencies. On the FERET database, where an affine normalization pre-processing was applied, the FBT algorithm outperformed only the PCA in a rank recognition test. However, it achieved performance comparable to state-of-the-art methods when evaluated by verification tests. These results indicate the high informative value of the polar frequency content of face images in relation to recognition and verification tasks, and that the Cartesian frequency content can complement information about the subjects' identity, but possibly only when the images are not pre-normalized. Possible implications for human face recognition are discussed. | Face Recognition Based on Polar Frequency Features | 2,970 |
We present a novel local-based face verification system whose components are analogous to those of biological systems. In the proposed system, after global registration and normalization, three eye regions are converted from the spatial to polar frequency domain by a Fourier-Bessel Transform. The resulting representations are embedded in a dissimilarity space, where each image is represented by its distance to all the other images. In this dissimilarity space a Pseudo-Fisher discriminator is built. ROC and equal error rate verification test results on the FERET database showed that the system performed at least as state-of-the-art methods and better than a system based on polar Fourier features. The local-based system is especially robust to facial expression and age variations, but sensitive to registration errors. | Face Verification in Polar Frequency Domain: a Biologically Motivated
Approach | 2,971 |
We present a method for automated segmentation of the vasculature in retinal images. The method produces segmentations by classifying each image pixel as vessel or non-vessel, based on the pixel's feature vector. Feature vectors are composed of the pixel's intensity and continuous two-dimensional Morlet wavelet transform responses taken at multiple scales. The Morlet wavelet is capable of tuning to specific frequencies, thus allowing noise filtering and vessel enhancement in a single step. We use a Bayesian classifier with class-conditional probability density functions (likelihoods) described as Gaussian mixtures, yielding a fast classification, while being able to model complex decision surfaces and compare its performance with the linear minimum squared error classifier. The probability distributions are estimated based on a training set of labeled pixels obtained from manual segmentations. The method's performance is evaluated on publicly available DRIVE and STARE databases of manually labeled non-mydriatic images. On the DRIVE database, it achieves an area under the receiver operating characteristic (ROC) curve of 0.9598, being slightly superior than that presented by the method of Staal et al. | Retinal Vessel Segmentation Using the 2-D Morlet Wavelet and Supervised
Classification | 2,972 |
In this paper, a decision support system for ship identification is presented. The system receives as input a silhouette of the vessel to be identified, previously extracted from a side view of the object. This view could have been acquired with imaging sensors operating at different spectral ranges (CCD, FLIR, image intensifier). The input silhouette is preprocessed and compared to those stored in a database, retrieving a small number of potential matches ranked by their similarity to the target silhouette. This set of potential matches is presented to the system operator, who makes the final ship identification. This system makes use of an evolved version of the Curvature Scale Space (CSS) representation. In the proposed approach, it is curvature extrema, instead of zero crossings, that are tracked during silhouette evolution, hence improving robustness and enabling to cope successfully with cases where the standard CCS representation is found to be unstable. Also, the use of local curvature was replaced with the more robust concept of lobe concavity, with significant additional gains in performance. Experimental results on actual operational imagery prove the excellent performance and robustness of the developed method. | A decision support system for ship identification based on the curvature
scale space representation | 2,973 |
Metal melting on release after explosion is a physical system far from quilibrium. A complete physical model of this system does not exist, because many interrelated effects have to be considered. General methodology needs to be developed so as to describe and understand physical phenomena involved. The high noise of the data, moving blur of images, the high degree of uncertainty due to the different types of sensors, and the information entangled and hidden inside the noisy images makes reasoning about the physical processes very difficult. Major problems include proper information extraction and the problem of reconstruction, as well as prediction of the missing data. In this paper, several techniques addressing the first problem are given, building the basis for tackling the second problem. | Understanding physics from interconnected data | 2,974 |
A novel algorithm for tunable compression to within the precision of reproduction targets, or storage, is proposed. The new algorithm is termed the `Perceptron Algorithm', which utilises simple existing concepts in a novel way, has multiple immediate commercial application aspects as well as it opens up a multitude of fronts in computational science and technology. The aims of this paper are to present the concepts underlying the algorithm, observations by its application to some example cases, and the identification of a multitude of potential areas of applications such as: image compression by orders of magnitude, signal compression including sound as well, image analysis in a multilayered detailed analysis, pattern recognition and matching and rapid database searching (e.g. face recognition), motion analysis, biomedical applications e.g. in MRI and CAT scan image analysis and compression, as well as hints on the link of these ideas to the way how biological memory might work leading to new points of view in neural computation. Commercial applications of immediate interest are the compression of images at the source (e.g. photographic equipment, scanners, satellite imaging systems), DVD film compression, pay-per-view downloads acceleration and many others identified in the present paper at its conclusion and future work section. | The Perceptron Algorithm: Image and Signal Decomposition, Compression,
and Analysis by Iterative Gaussian Blurring | 2,975 |
This short article presents an alternative view of high resolution imaging from various sources with the aim of the discovery of potential sites of archaeological importance, or sites that exhibit `anomalies' such that they may merit closer inspection and analysis. It is conjectured, and to a certain extent demonstrated here, that it is possible for advanced civilizations to factor in erosion by natural processes into a large scale design so that main features be preserved even with the passage of millions of years. Alternatively viewed, even without such intent embedded in a design left for posterity, it is possible that a gigantic construction may naturally decay in such a way that even cataclysmic (massive) events may leave sufficient information intact with the passage of time, provided one changes the point of view from high resolution images to enhanced blurred renderings of the sites in question. | The `Face on Mars': a photographic approach for the search of signs of
past civilizations from a macroscopic point of view, factoring long-term
erosion in image reconstruction | 2,976 |
A novel algorithm is proposed for segmenting an image into multiple levels using its mean and variance. Starting from the extreme pixel values at both ends of the histogram plot, the algorithm is applied recursively on sub-ranges computed from the previous step, so as to find a threshold level and a new sub-range for the next step, until no significant improvement in image quality can be achieved. The method makes use of the fact that a number of distributions tend towards Dirac delta function, peaking at the mean, in the limiting condition of vanishing variance. The procedure naturally provides for variable size segmentation with bigger blocks near the extreme pixel values and finer divisions around the mean or other chosen value for better visualization. Experiments on a variety of images show that the new algorithm effectively segments the image in computationally very less time. | Multilevel Thresholding for Image Segmentation through a Fast
Statistical Recursive Algorithm | 2,977 |
We present an algorithm that enables one to perform locally adaptive block thresholding, while maintaining image continuity. Images are divided into sub-images based some standard image attributes and thresholding technique is employed over the sub-images. The present algorithm makes use of the thresholds of neighboring sub-images to calculate a range of values. The image continuity is taken care by choosing the threshold of the sub-image under consideration to lie within the above range. After examining the average range values for various sub-image sizes of a variety of images, it was found that the range of acceptable threshold values is substantially high, justifying our assumption of exploiting the freedom of range for bringing out local details. | Locally Adaptive Block Thresholding Method with Continuity Constraint | 2,978 |
This communication describes a representation of images as a set of edges characterized by their position and orientation. This representation allows the comparison of two images and the computation of their similarity. The first step in this computation of similarity is the seach of a geometrical basis of the two dimensional space where the two images are represented simultaneously after transformation of one of them. Presently, this simultaneous representation takes into account a shift and a scaling ; it may be extended to rotations or other global geometrical transformations. An elementary probabilistic computation shows that a sufficient but not excessive number of trials (a few tens) ensures that the exhibition of this common basis is guaranteed in spite of possible errors in the detection of edges. When this first step is performed, the search of similarity between the two images reduces to counting the coincidence of edges in the two images. The approach may be applied to many problems of pattern matching ; it was checked on face recognition. | Matching Edges in Images ; Application to Face Recognition | 2,979 |
In this paper, we focus on Fourier analysis and holographic transforms for signal representation. For instance, in the case of image processing, the holographic representation has the property that an arbitrary portion of the transformed image enables reconstruction of the whole image with details missing. We focus on holographic representation defined through the Fourier Transforms. Thus, We firstly review some results in Fourier transform and Fourier series. Next, we review the Discrete Holographic Fourier Transform (DHFT) for image representation. Then, we describe the contributions of our work. We show a simple scheme for progressive transmission based on the DHFT. Next, we propose the Continuous Holographic Fourier Transform (CHFT) and discuss some theoretical aspects of it for 1D signals. Finally, some testes are presented in the experimental results | Fourier Analysis and Holographic Representations of 1D and 2D Signals | 2,980 |
Feature extraction and matching are among central problems of computer vision. It is inefficent to search features over all locations and scales. Neurophysiological evidence shows that to locate objects in a digital image the human visual system employs visual attention to a specific object while ignoring others. The brain also has a mechanism to search from coarse to fine. In this paper, we present a feature extractor and an associated hierarchical searching model to simulate such processes. With the hierarchical representation of the object, coarse scanning is done through the matching of the larger scale and precise localization is conducted through the matching of the smaller scale. Experimental results justify the proposed model in its effectiveness and efficiency to localize features. | Biologically Inspired Hierarchical Model for Feature Extraction and
Localization | 2,981 |
In this article we propose a novel face recognition method based on Principal Component Analysis (PCA) and Log-Gabor filters. The main advantages of the proposed method are its simple implementation, training, and very high recognition accuracy. For recognition experiments we used 5151 face images of 1311 persons from different sets of the FERET and AR databases that allow to analyze how recognition accuracy is affected by the change of facial expressions, illumination, and aging. Recognition experiments with the FERET database (containing photographs of 1196 persons) showed that our method can achieve maximal 97-98% first one recognition rate and 0.3-0.4% Equal Error Rate. The experiments also showed that the accuracy of our method is less affected by eye location errors and used image normalization method than of traditional PCA -based recognition method. | Face Recognition using Principal Component Analysis and Log-Gabor
Filters | 2,982 |
In this article we propose a method for the recognition of faces with different facial expressions. For recognition we extract feature vectors by using log-Gabor filters of multiple orientations and scales. Using sliding window algorithm and variances -based masking these features are extracted at image regions that are less affected by the changes of facial expressions. Extracted features are passed to the Principal Component Analysis (PCA) -based recognition method. The results of face recognition experiments using expression variant faces showed that the proposed method could achieve higher recognition accuracy than many other methods. For development and testing we used facial images from the AR and FERET databases. Using facial photographs of more than one thousand persons from the FERET database the proposed method achieved 96.6-98.9% first one recognition rate and 0.2-0.6% Equal Error Rate (EER). | Recognition of expression variant faces using masked log-Gabor features
and Principal Component Analysis | 2,983 |
Regularization functionals that lower level set boundary length when used with L^1 fidelity functionals on signal de-noising on images create artifacts. These are (i) rounding of corners, (ii) shrinking of radii, (iii) shrinking of cusps, and (iv) non-smoothing of staircasing. Regularity functionals based upon total curvature of level set boundaries do not create artifacts (i) and (ii). An adjusted fidelity term based on the flat norm on the current (a distributional graph) representing the density of curvature of level sets boundaries can minimize (iii) by weighting the position of a cusp. A regularity term to eliminate staircasing can be based upon the mass of the current representing the graph of an image function or its second derivatives. Densities on the Grassmann bundle of the Grassmann bundle of the ambient space of the graph can be used to identify patterns, textures, occlusion and lines. | Notes on Geometric Measure Theory Applications to Image Processing;
De-noising, Segmentation, Pattern, Texture, Lines, Gestalt and Occlusion | 2,984 |
Raster images can have a range of various distortions connected to their raster structure. Upsampling them might in effect substantially yield the raster structure of the original image, known as aliasing. The upsampling itself may introduce aliasing into the upsampled image as well. The presented method attempts to remove the aliasing using frequency filters based on the discrete fast Fourier transform, and applied directionally in certain regions placed along the edges in the image. As opposed to some anisotropic smoothing methods, the presented algorithm aims to selectively reduce only the aliasing, preserving the sharpness of image details. The method can be used as a post--processing filter along with various upsampling algorithms. It was experimentally shown that the method can improve the visual quality of the upsampled images. | An effective edge--directed frequency filter for removal of aliasing in
upsampled images | 2,985 |
In this paper, we are interested in the application to video segmentation of the discrete shape optimization problem involving the shape weighted perimeter and an additional term depending on a parameter. Based on recent works and in particular the one of Darbon and Sigelle, we justify the equivalence of the shape optimization problem and a weighted total variation regularization. For solving this problem, we adapt the projection algorithm proposed recently for solving the basic TV regularization problem. Another solution to the shape optimization investigated here is the graph cut technique. Both methods have the advantage to lead to a global minimum. Since we can distinguish moving objects from static elements of a scene by analyzing norm of the optical flow vectors, we choose the optical flow norm as initial data. In order to have the contour as close as possible to an edge in the image, we use a classical edge detector function as the weight of the weighted total variation. This model has been used in one of our former works. We also apply the same methods to a video segmentation model used by Jehan-Besson, Barlaud and Aubert. In this case, only standard perimeter is incorporated in the shape functional. We also propose another way for finding moving objects by using an a contrario detection of objects on the image obtained by solving the Rudin-Osher-Fatemi Total Variation regularization problem.We can notice the segmentation can be associated to a level set in the former methods. | Total Variation Minimization and Graph Cuts for Moving Objects
Segmentation | 2,986 |
We present conditional expression (CE) for finding blurs convolved in given images. The CE is given in terms of the zero-values of the blurs evaluated at multi-point. The CE can detect multiple blur all at once. We illustrate the multiple blur-detection by using a test image. | Conditional Expressions for Blind Deconvolution: Multi-point form | 2,987 |
We developed novel conditional expressions (CEs) for Lane and Bates' blind deconvolution. The CEs are given in term of the derivatives of the zero-values of the z-transform of given images. The CEs make it possible to automatically detect multiple blur convolved in the given images all at once without performing any analysis of the zero-sheets of the given images. We illustrate the multiple blur-detection by the CEs for a model image | Conditional Expressions for Blind Deconvolution: Derivative form | 2,988 |
In this paper, we propose a global method for estimating the motion of a camera which films a static scene. Our approach is direct, fast and robust, and deals with adjacent frames of a sequence. It is based on a quadratic approximation of the deformation between two images, in the case of a scene with constant depth in the camera coordinate system. This condition is very restrictive but we show that provided translation and depth inverse variations are small enough, the error on optical flow involved by the approximation of depths by a constant is small. In this context, we propose a new model of camera motion, that allows to separate the image deformation in a similarity and a ``purely'' projective application, due to change of optical axis direction. This model leads to a quadratic approximation of image deformation that we estimate with an M-estimator; we can immediatly deduce camera motion parameters. | Camera motion estimation through planar deformation determination | 2,989 |
Many image processing problems involve identifying the region in the image domain occupied by a given entity in the scene. Automatic solution of these problems requires models that incorporate significant prior knowledge about the shape of the region. Many methods for including such knowledge run into difficulties when the topology of the region is unknown a priori, for example when the entity is composed of an unknown number of similar objects. Higher-order active contours (HOACs) represent one method for the modelling of non-trivial prior knowledge about shape without necessarily constraining region topology, via the inclusion of non-local interactions between region boundary points in the energy defining the model. The case of an unknown number of circular objects arises in a number of domains, e.g. medical, biological, nanotechnological, and remote sensing imagery. Regions composed of an a priori unknown number of circles may be referred to as a `gas of circles'. In this report, we present a HOAC model of a `gas of circles'. In order to guarantee stable circles, we conduct a stability analysis via a functional Taylor expansion of the HOAC energy around a circular shape. This analysis fixes one of the model parameters in terms of the others and constrains the rest. In conjunction with a suitable likelihood energy, we apply the model to the extraction of tree crowns from aerial imagery, and show that the new model outperforms other techniques. | A higher-order active contour model of a `gas of circles' and its
application to tree crown extraction | 2,990 |
Irregular pyramids are made of a stack of successively reduced graphs embedded in the plane. Such pyramids are used within the segmentation framework to encode a hierarchy of partitions. The different graph models used within the irregular pyramid framework encode different types of relationships between regions. This paper compares different graph models used within the irregular pyramid framework according to a set of relationships between regions. We also define a new algorithm based on a pyramid of combinatorial maps which allows to determine if one region contains the other using only local calculus. | Contains and Inside relationships within combinatorial Pyramids | 2,991 |
The aim of this study is to detect man-made cartographic objects in high-resolution satellite images. New generation satellites offer a sub-metric spatial resolution, in which it is possible (and necessary) to develop methods at object level rather than at pixel level, and to exploit structural features of objects. With this aim, a method to generate structural object models from manually segmented images has been developed. To generate the model from non-segmented images, extraction of the objects from the sample images is required. A hybrid method of extraction (both in terms of input sources and segmentation algorithms) is proposed: A region based segmentation is applied on a 10 meter resolution multi-spectral image. The result is used as marker in a "marker-controlled watershed method using edges" on a 2.5 meter resolution panchromatic image. Very promising results have been obtained even on images where the limits of the target objects are not apparent. | Extraction of cartographic objects in high resolution satellite images
for object model generation | 2,992 |
There is a huge amount of historical documents in libraries and in various National Archives that have not been exploited electronically. Although automatic reading of complete pages remains, in most cases, a long-term objective, tasks such as word spotting, text/image alignment, authentication and extraction of specific fields are in use today. For all these tasks, a major step is document segmentation into text lines. Because of the low quality and the complexity of these documents (background noise, artifacts due to aging, interfering lines),automatic text line segmentation remains an open research field. The objective of this paper is to present a survey of existing methods, developed during the last decade, and dedicated to documents of historical interest. | Text Line Segmentation of Historical Documents: a Survey | 2,993 |
We propose a new algorithm to the problem of polygonal curve approximation based on a multiresolution approach. This algorithm is suboptimal but still maintains some optimality between successive levels of resolution using dynamic programming. We show theoretically and experimentally that this algorithm has a linear complexity in time and space. We experimentally compare the outcomes of our algorithm to the optimal "full search" dynamic programming solution and finally to classical merge and split approaches. The experimental evaluations confirm the theoretical derivations and show that the proposed approach evaluated on 2D coastal maps either show a lower time complexity or provide polygonal approximations closer to the input discrete curves. | Multiresolution Approximation of Polygonal Curves in Linear Complexity | 2,994 |
This paper presents deformable templates as a tool for segmentation and localization of biological structures in medical images. Structures are represented by a prototype template, combined with a parametric warp mapping used to deform the original shape. The localization procedure is achieved using a multi-stage, multi-resolution algorithm de-signed to reduce computational complexity and time. The algorithm initially identifies regions in the image most likely to contain the desired objects and then examines these regions at progressively increasing resolutions. The final stage of the algorithm involves warping the prototype template to match the localized objects. The algorithm is presented along with the results of four example applications using MRI, x-ray and ultrasound images. | Medical Image Segmentation and Localization using Deformable Templates | 2,995 |
Nuclear medicine (NM) images inherently suffer from large amounts of noise and blur. The purpose of this research is to reduce the noise and blur while maintaining image integrity for improved diagnosis. The proposed solution is to increase image quality after the standard pre- and post-processing undertaken by a gamma camera system. Mean Field Annealing (MFA) is the image processing technique used in this research. It is a computational iterative technique that makes use of the Point Spread Function (PSF) and the noise associated with the NM image. MFA is applied to NM images with the objective of reducing noise while not compromising edge integrity. Using a sharpening filter as a post-processing technique (after MFA) yields image enhancement of planar NM images. | Enhancement of Noisy Planar Nuclear Medicine Images using Mean Field
Annealing | 2,996 |
This paper explores a comparative study of both the linear and kernel implementations of three of the most popular Appearance-based Face Recognition projection classes, these being the methodologies of Principal Component Analysis, Linear Discriminant Analysis and Independent Component Analysis. The experimental procedure provides a platform of equal working conditions and examines the ten algorithms in the categories of expression, illumination, occlusion and temporal delay. The results are then evaluated based on a sequential combination of assessment tools that facilitate both intuitive and statistical decisiveness among the intra and interclass comparisons. The best categorical algorithms are then incorporated into a hybrid methodology, where the advantageous effects of fusion strategies are considered. | An Independent Evaluation of Subspace Face Recognition Algorithms | 2,997 |
Subtraction of aligned images is a means to assess changes in a wide variety of clinical applications. In this paper we explore the information theoretical origin of Mutual Information (MI), which is based on Shannon's entropy.However, the interpretation of standard MI registration as a communication channel suggests that MI is too restrictive a criterion. In this paper the concept of Mutual Information (MI) is extended to (Normalized) Focussed Mutual Information (FMI) to incorporate prior knowledge to overcome some shortcomings of MI. We use this to develop new methodologies to successfully address specific registration problems, the follow-up of dental restorations, cephalometry, and the monitoring of implants. | MI image registration using prior knowledge | 2,998 |
This article describes the implementation of a system designed to automatically detect the presence of pulmonary embolism in lung scans. These images are firstly segmented, before alignment and feature extraction using PCA. The neural network was trained using the Hybrid Monte Carlo method, resulting in a committee of 250 neural networks and good results are obtained. | Automatic Detection of Pulmonary Embolism using Computational
Intelligence | 2,999 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.