id
stringlengths
9
10
submitter
stringlengths
1
64
authors
stringlengths
4
20.7k
title
stringlengths
4
246
comments
stringlengths
1
523
journal-ref
stringlengths
4
404
doi
stringlengths
11
153
report-no
stringlengths
2
254
categories
stringlengths
5
98
license
stringclasses
9 values
orig_abstract
stringlengths
14
3.35k
versions
listlengths
1
60
update_date
stringlengths
10
10
authors_parsed
listlengths
1
1.35k
abstract
stringlengths
11
3.34k
1207.4121
Fabio Gagliardi Cozman
Fabio Gagliardi Cozman, Cassio Polpo de Campos, Jaime Ide, Jose Carlos Ferreira da Rocha
Propositional and Relational Bayesian Networks Associated with Imprecise and Qualitative Probabilistic Assesments
Appears in Proceedings of the Twentieth Conference on Uncertainty in Artificial Intelligence (UAI2004)
null
null
UAI-P-2004-PG-104-111
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper investigates a representation language with flexibility inspired by probabilistic logic and compactness inspired by relational Bayesian networks. The goal is to handle propositional and first-order constructs together with precise, imprecise, indeterminate and qualitative probabilistic assessments. The paper shows how this can be achieved through the theory of credal networks. New exact and approximate inference algorithms based on multilinear programming and iterated/loopy propagation of interval probabilities are presented; their superior performance, compared to existing ones, is shown empirically.
[ { "created": "Wed, 11 Jul 2012 14:45:39 GMT", "version": "v1" } ]
2012-07-19
[ [ "Cozman", "Fabio Gagliardi", "" ], [ "de Campos", "Cassio Polpo", "" ], [ "Ide", "Jaime", "" ], [ "da Rocha", "Jose Carlos Ferreira", "" ] ]
This paper investigates a representation language with flexibility inspired by probabilistic logic and compactness inspired by relational Bayesian networks. The goal is to handle propositional and first-order constructs together with precise, imprecise, indeterminate and qualitative probabilistic assessments. The paper shows how this can be achieved through the theory of credal networks. New exact and approximate inference algorithms based on multilinear programming and iterated/loopy propagation of interval probabilities are presented; their superior performance, compared to existing ones, is shown empirically.
2010.11425
Abhimanyu Dubey
Abhimanyu Dubey and Alex Pentland
Differentially-Private Federated Linear Bandits
22 pages. Camera-ready for NeurIPS 2020
null
null
null
cs.LG cs.CR cs.MA stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The rapid proliferation of decentralized learning systems mandates the need for differentially-private cooperative learning. In this paper, we study this in context of the contextual linear bandit: we consider a collection of agents cooperating to solve a common contextual bandit, while ensuring that their communication remains private. For this problem, we devise \textsc{FedUCB}, a multiagent private algorithm for both centralized and decentralized (peer-to-peer) federated learning. We provide a rigorous technical analysis of its utility in terms of regret, improving several results in cooperative bandit learning, and provide rigorous privacy guarantees as well. Our algorithms provide competitive performance both in terms of pseudoregret bounds and empirical benchmark performance in various multi-agent settings.
[ { "created": "Thu, 22 Oct 2020 03:58:39 GMT", "version": "v1" } ]
2020-10-23
[ [ "Dubey", "Abhimanyu", "" ], [ "Pentland", "Alex", "" ] ]
The rapid proliferation of decentralized learning systems mandates the need for differentially-private cooperative learning. In this paper, we study this in context of the contextual linear bandit: we consider a collection of agents cooperating to solve a common contextual bandit, while ensuring that their communication remains private. For this problem, we devise \textsc{FedUCB}, a multiagent private algorithm for both centralized and decentralized (peer-to-peer) federated learning. We provide a rigorous technical analysis of its utility in terms of regret, improving several results in cooperative bandit learning, and provide rigorous privacy guarantees as well. Our algorithms provide competitive performance both in terms of pseudoregret bounds and empirical benchmark performance in various multi-agent settings.
1404.1462
Pallavi Sudhakar
Pallavi.V.S, Dr. Rukmani Devi.D
Design of a High Speed FPGA-Based Classifier for Efficient Packet Classification
6 pages, 6 figures, "Published with International Journal of Computer Trends and Technology (IJCTT)", "National Conference on Modern Electronics and Signal Processing (2014)- Velammal Engineering College", "Recent Trends in Information Technology (2014)- R.M.K College of Engineering and Technology"
Pallavi.V.S, Dr.Rukmani Devi.D Article: Design of a High Speed FPGA-Based Classifier for Efficient Packet Classification. International Journal of Computer Trends and Technology(IJCTT) 9(3):123-128,Mar 2014
10.14445/22312803/IJCTT-V9P126
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Packet classification is a vital and complicated task as the processing of packets should be done at a specified line speed. In order to classify a packet as belonging to a particular flow or set of flows, network nodes must perform a search over a set of filters using multiple fields of the packet as the search key. Hence the matching of packets should be much faster and simpler for quick processing and classification. A hardware accelerator or a classifier has been proposed here using a modified version of the HyperCuts packet classification algorithm. A new pre-cutting process has been implemented to reduce the memory size to fit in an FPGA. This classifier can classify packets with high speed and with a power consumption factor of less than 3W. This methodology removes the need for floating point division to be performed by replacing the region compaction scheme of HyperCuts by pre-cutting, while classifying the packets and concentrates on classifying the packets at the core of the network.
[ { "created": "Sat, 5 Apr 2014 11:06:40 GMT", "version": "v1" } ]
2014-04-08
[ [ "S", "Pallavi. V.", "" ], [ "D", "Dr. Rukmani Devi.", "" ] ]
Packet classification is a vital and complicated task as the processing of packets should be done at a specified line speed. In order to classify a packet as belonging to a particular flow or set of flows, network nodes must perform a search over a set of filters using multiple fields of the packet as the search key. Hence the matching of packets should be much faster and simpler for quick processing and classification. A hardware accelerator or a classifier has been proposed here using a modified version of the HyperCuts packet classification algorithm. A new pre-cutting process has been implemented to reduce the memory size to fit in an FPGA. This classifier can classify packets with high speed and with a power consumption factor of less than 3W. This methodology removes the need for floating point division to be performed by replacing the region compaction scheme of HyperCuts by pre-cutting, while classifying the packets and concentrates on classifying the packets at the core of the network.
2307.03741
Vladan Stojni\'c
Vladan Stojni\'c, Zakaria Laskar, Giorgos Tolias
Training Ensembles with Inliers and Outliers for Semi-supervised Active Learning
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep active learning in the presence of outlier examples poses a realistic yet challenging scenario. Acquiring unlabeled data for annotation requires a delicate balance between avoiding outliers to conserve the annotation budget and prioritizing useful inlier examples for effective training. In this work, we present an approach that leverages three highly synergistic components, which are identified as key ingredients: joint classifier training with inliers and outliers, semi-supervised learning through pseudo-labeling, and model ensembling. Our work demonstrates that ensembling significantly enhances the accuracy of pseudo-labeling and improves the quality of data acquisition. By enabling semi-supervision through the joint training process, where outliers are properly handled, we observe a substantial boost in classifier accuracy through the use of all available unlabeled examples. Notably, we reveal that the integration of joint training renders explicit outlier detection unnecessary; a conventional component for acquisition in prior work. The three key components align seamlessly with numerous existing approaches. Through empirical evaluations, we showcase that their combined use leads to a performance increase. Remarkably, despite its simplicity, our proposed approach outperforms all other methods in terms of performance. Code: https://github.com/vladan-stojnic/active-outliers
[ { "created": "Fri, 7 Jul 2023 17:50:07 GMT", "version": "v1" } ]
2023-07-10
[ [ "Stojnić", "Vladan", "" ], [ "Laskar", "Zakaria", "" ], [ "Tolias", "Giorgos", "" ] ]
Deep active learning in the presence of outlier examples poses a realistic yet challenging scenario. Acquiring unlabeled data for annotation requires a delicate balance between avoiding outliers to conserve the annotation budget and prioritizing useful inlier examples for effective training. In this work, we present an approach that leverages three highly synergistic components, which are identified as key ingredients: joint classifier training with inliers and outliers, semi-supervised learning through pseudo-labeling, and model ensembling. Our work demonstrates that ensembling significantly enhances the accuracy of pseudo-labeling and improves the quality of data acquisition. By enabling semi-supervision through the joint training process, where outliers are properly handled, we observe a substantial boost in classifier accuracy through the use of all available unlabeled examples. Notably, we reveal that the integration of joint training renders explicit outlier detection unnecessary; a conventional component for acquisition in prior work. The three key components align seamlessly with numerous existing approaches. Through empirical evaluations, we showcase that their combined use leads to a performance increase. Remarkably, despite its simplicity, our proposed approach outperforms all other methods in terms of performance. Code: https://github.com/vladan-stojnic/active-outliers
2408.05518
Hongsheng Qin
Zhengang Lu, Hongsheng Qin, Jing Li, Ming Sun and Jiubin Tan
Long working distance portable smartphone microscopy for metallic mesh defect detection
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Metallic mesh is a transparent electromagnetic shielding film with a fine metal line structure. However, it can develop defects that affect the optoelectronic performance whether in the production preparation or in actual use. The development of in-situ non-destructive testing (NDT) devices for metallic mesh requires long working distances, reflective optical path design, and miniaturization. To address the limitations of existing smartphone microscopes, which feature short working distances and inadequate transmission imaging for industrial in-situ inspection, we propose a novel long-working distance reflective smartphone microscopy system (LD-RSM). LD-RSM builds a 4f optical imaging system with external optical components and a smartphone, utilizing a beam splitter to achieve reflective imaging with the illumination system and imaging system on the same side of the sample. It achieves an optical resolution of 4.92$\mu$m and a working distance of up to 22.23 mm. Additionally, we introduce a dual prior weighted Robust Principal Component Analysis (DW-RPCA) for defect detection. This approach leverages spectral filter fusion and Hough transform to model different defect types, enhancing the accuracy and efficiency of defect identification. Coupled with an optimized threshold segmentation algorithm, DW-RPCA method achieves a pixel-level accuracy of 84.8%. Our work showcases strong potential for growth in the field of in-situ on-line inspection of industrial products.
[ { "created": "Sat, 10 Aug 2024 11:02:03 GMT", "version": "v1" }, { "created": "Tue, 13 Aug 2024 05:16:07 GMT", "version": "v2" } ]
2024-08-14
[ [ "Lu", "Zhengang", "" ], [ "Qin", "Hongsheng", "" ], [ "Li", "Jing", "" ], [ "Sun", "Ming", "" ], [ "Tan", "Jiubin", "" ] ]
Metallic mesh is a transparent electromagnetic shielding film with a fine metal line structure. However, it can develop defects that affect the optoelectronic performance whether in the production preparation or in actual use. The development of in-situ non-destructive testing (NDT) devices for metallic mesh requires long working distances, reflective optical path design, and miniaturization. To address the limitations of existing smartphone microscopes, which feature short working distances and inadequate transmission imaging for industrial in-situ inspection, we propose a novel long-working distance reflective smartphone microscopy system (LD-RSM). LD-RSM builds a 4f optical imaging system with external optical components and a smartphone, utilizing a beam splitter to achieve reflective imaging with the illumination system and imaging system on the same side of the sample. It achieves an optical resolution of 4.92$\mu$m and a working distance of up to 22.23 mm. Additionally, we introduce a dual prior weighted Robust Principal Component Analysis (DW-RPCA) for defect detection. This approach leverages spectral filter fusion and Hough transform to model different defect types, enhancing the accuracy and efficiency of defect identification. Coupled with an optimized threshold segmentation algorithm, DW-RPCA method achieves a pixel-level accuracy of 84.8%. Our work showcases strong potential for growth in the field of in-situ on-line inspection of industrial products.
2205.07854
Haoteng Tang
Haoteng Tang, Xiyao Fu, Lei Guo, Yalin Wang, Scott Mackin, Olusola Ajilore, Alex Leow, Paul Thompson, Heng Huang, Liang Zhan
Functional2Structural: Cross-Modality Brain Networks Representation Learning
null
null
null
null
cs.LG cs.AI cs.CV eess.IV q-bio.NC
http://creativecommons.org/licenses/by/4.0/
MRI-based modeling of brain networks has been widely used to understand functional and structural interactions and connections among brain regions, and factors that affect them, such as brain development and disease. Graph mining on brain networks may facilitate the discovery of novel biomarkers for clinical phenotypes and neurodegenerative diseases. Since brain networks derived from functional and structural MRI describe the brain topology from different perspectives, exploring a representation that combines these cross-modality brain networks is non-trivial. Most current studies aim to extract a fused representation of the two types of brain network by projecting the structural network to the functional counterpart. Since the functional network is dynamic and the structural network is static, mapping a static object to a dynamic object is suboptimal. However, mapping in the opposite direction is not feasible due to the non-negativity requirement of current graph learning techniques. Here, we propose a novel graph learning framework, known as Deep Signed Brain Networks (DSBN), with a signed graph encoder that, from an opposite perspective, learns the cross-modality representations by projecting the functional network to the structural counterpart. We validate our framework on clinical phenotype and neurodegenerative disease prediction tasks using two independent, publicly available datasets (HCP and OASIS). The experimental results clearly demonstrate the advantages of our model compared to several state-of-the-art methods.
[ { "created": "Fri, 6 May 2022 03:45:36 GMT", "version": "v1" } ]
2022-05-18
[ [ "Tang", "Haoteng", "" ], [ "Fu", "Xiyao", "" ], [ "Guo", "Lei", "" ], [ "Wang", "Yalin", "" ], [ "Mackin", "Scott", "" ], [ "Ajilore", "Olusola", "" ], [ "Leow", "Alex", "" ], [ "Thompson", "Paul", "" ], [ "Huang", "Heng", "" ], [ "Zhan", "Liang", "" ] ]
MRI-based modeling of brain networks has been widely used to understand functional and structural interactions and connections among brain regions, and factors that affect them, such as brain development and disease. Graph mining on brain networks may facilitate the discovery of novel biomarkers for clinical phenotypes and neurodegenerative diseases. Since brain networks derived from functional and structural MRI describe the brain topology from different perspectives, exploring a representation that combines these cross-modality brain networks is non-trivial. Most current studies aim to extract a fused representation of the two types of brain network by projecting the structural network to the functional counterpart. Since the functional network is dynamic and the structural network is static, mapping a static object to a dynamic object is suboptimal. However, mapping in the opposite direction is not feasible due to the non-negativity requirement of current graph learning techniques. Here, we propose a novel graph learning framework, known as Deep Signed Brain Networks (DSBN), with a signed graph encoder that, from an opposite perspective, learns the cross-modality representations by projecting the functional network to the structural counterpart. We validate our framework on clinical phenotype and neurodegenerative disease prediction tasks using two independent, publicly available datasets (HCP and OASIS). The experimental results clearly demonstrate the advantages of our model compared to several state-of-the-art methods.
1210.5041
Thomas Maugey
Thomas Maugey, Ismael Daribo, Gene Cheung, Pascal Frossard
Navigation domain representation for interactive multiview imaging
null
null
10.1109/TIP.2013.2270183
null
cs.MM cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Enabling users to interactively navigate through different viewpoints of a static scene is a new interesting functionality in 3D streaming systems. While it opens exciting perspectives towards rich multimedia applications, it requires the design of novel representations and coding techniques in order to solve the new challenges imposed by interactive navigation. Interactivity clearly brings new design constraints: the encoder is unaware of the exact decoding process, while the decoder has to reconstruct information from incomplete subsets of data since the server can generally not transmit images for all possible viewpoints due to resource constrains. In this paper, we propose a novel multiview data representation that permits to satisfy bandwidth and storage constraints in an interactive multiview streaming system. In particular, we partition the multiview navigation domain into segments, each of which is described by a reference image and some auxiliary information. The auxiliary information enables the client to recreate any viewpoint in the navigation segment via view synthesis. The decoder is then able to navigate freely in the segment without further data request to the server; it requests additional data only when it moves to a different segment. We discuss the benefits of this novel representation in interactive navigation systems and further propose a method to optimize the partitioning of the navigation domain into independent segments, under bandwidth and storage constraints. Experimental results confirm the potential of the proposed representation; namely, our system leads to similar compression performance as classical inter-view coding, while it provides the high level of flexibility that is required for interactive streaming. Hence, our new framework represents a promising solution for 3D data representation in novel interactive multimedia services.
[ { "created": "Thu, 18 Oct 2012 07:41:17 GMT", "version": "v1" }, { "created": "Mon, 17 Jun 2013 09:32:50 GMT", "version": "v2" } ]
2015-06-11
[ [ "Maugey", "Thomas", "" ], [ "Daribo", "Ismael", "" ], [ "Cheung", "Gene", "" ], [ "Frossard", "Pascal", "" ] ]
Enabling users to interactively navigate through different viewpoints of a static scene is a new interesting functionality in 3D streaming systems. While it opens exciting perspectives towards rich multimedia applications, it requires the design of novel representations and coding techniques in order to solve the new challenges imposed by interactive navigation. Interactivity clearly brings new design constraints: the encoder is unaware of the exact decoding process, while the decoder has to reconstruct information from incomplete subsets of data since the server can generally not transmit images for all possible viewpoints due to resource constrains. In this paper, we propose a novel multiview data representation that permits to satisfy bandwidth and storage constraints in an interactive multiview streaming system. In particular, we partition the multiview navigation domain into segments, each of which is described by a reference image and some auxiliary information. The auxiliary information enables the client to recreate any viewpoint in the navigation segment via view synthesis. The decoder is then able to navigate freely in the segment without further data request to the server; it requests additional data only when it moves to a different segment. We discuss the benefits of this novel representation in interactive navigation systems and further propose a method to optimize the partitioning of the navigation domain into independent segments, under bandwidth and storage constraints. Experimental results confirm the potential of the proposed representation; namely, our system leads to similar compression performance as classical inter-view coding, while it provides the high level of flexibility that is required for interactive streaming. Hence, our new framework represents a promising solution for 3D data representation in novel interactive multimedia services.
1904.11327
Saverio Giallorenzo
Saverio Giallorenzo, Fabrizio Montesi, Larisa Safina, Stefano Pio Zingaro
Ephemeral Data Handling in Microservices - Technical Report
null
null
null
null
cs.PL cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In modern application areas for software systems --- like eHealth, the Internet-of-Things, and Edge Computing --- data is encoded in heterogeneous, tree-shaped data-formats, it must be processed in real-time, and it must be ephemeral, i.e., not persist in the system. While it is preferable to use a query language to express complex data-handling logic, their typical execution engine, a database external from the main application, is unfit in scenarios of ephemeral data-handling. A better option is represented by integrated query frameworks, which benefit from existing development support tools (e.g., syntax and type checkers) and execute within the application memory. In this paper, we propose one such framework that, for the first time, targets tree-shaped, document-oriented queries. We formalise an instantiation of MQuery, a sound variant of the widely-used MongoDB query language, which we implemented in the Jolie language. Jolie programs are microservices, the building blocks of modern software systems. Moreover, since Jolie supports native tree data-structures and automatic management of heterogeneous data-encodings, we can provide a uniform way to use MQuery on any data-format supported by the language. We present a non-trivial use case from eHealth, use it to concretely evaluate our model, and to illustrate our formalism.
[ { "created": "Thu, 25 Apr 2019 13:31:33 GMT", "version": "v1" } ]
2019-04-26
[ [ "Giallorenzo", "Saverio", "" ], [ "Montesi", "Fabrizio", "" ], [ "Safina", "Larisa", "" ], [ "Zingaro", "Stefano Pio", "" ] ]
In modern application areas for software systems --- like eHealth, the Internet-of-Things, and Edge Computing --- data is encoded in heterogeneous, tree-shaped data-formats, it must be processed in real-time, and it must be ephemeral, i.e., not persist in the system. While it is preferable to use a query language to express complex data-handling logic, their typical execution engine, a database external from the main application, is unfit in scenarios of ephemeral data-handling. A better option is represented by integrated query frameworks, which benefit from existing development support tools (e.g., syntax and type checkers) and execute within the application memory. In this paper, we propose one such framework that, for the first time, targets tree-shaped, document-oriented queries. We formalise an instantiation of MQuery, a sound variant of the widely-used MongoDB query language, which we implemented in the Jolie language. Jolie programs are microservices, the building blocks of modern software systems. Moreover, since Jolie supports native tree data-structures and automatic management of heterogeneous data-encodings, we can provide a uniform way to use MQuery on any data-format supported by the language. We present a non-trivial use case from eHealth, use it to concretely evaluate our model, and to illustrate our formalism.
2404.12704
Haoyu Sun
Jiazhu Dai, Haoyu Sun
A Clean-graph Backdoor Attack against Graph Convolutional Networks with Poisoned Label Only
null
null
null
null
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Graph Convolutional Networks (GCNs) have shown excellent performance in dealing with various graph structures such as node classification, graph classification and other tasks. However,recent studies have shown that GCNs are vulnerable to a novel threat known as backdoor attacks. However, all existing backdoor attacks in the graph domain require modifying the training samples to accomplish the backdoor injection, which may not be practical in many realistic scenarios where adversaries have no access to modify the training samples and may leads to the backdoor attack being detected easily. In order to explore the backdoor vulnerability of GCNs and create a more practical and stealthy backdoor attack method, this paper proposes a clean-graph backdoor attack against GCNs (CBAG) in the node classification task,which only poisons the training labels without any modification to the training samples, revealing that GCNs have this security vulnerability. Specifically, CBAG designs a new trigger exploration method to find important feature dimensions as the trigger patterns to improve the attack performance. By poisoning the training labels, a hidden backdoor is injected into the GCNs model. Experimental results show that our clean graph backdoor can achieve 99% attack success rate while maintaining the functionality of the GCNs model on benign samples.
[ { "created": "Fri, 19 Apr 2024 08:21:54 GMT", "version": "v1" } ]
2024-04-22
[ [ "Dai", "Jiazhu", "" ], [ "Sun", "Haoyu", "" ] ]
Graph Convolutional Networks (GCNs) have shown excellent performance in dealing with various graph structures such as node classification, graph classification and other tasks. However,recent studies have shown that GCNs are vulnerable to a novel threat known as backdoor attacks. However, all existing backdoor attacks in the graph domain require modifying the training samples to accomplish the backdoor injection, which may not be practical in many realistic scenarios where adversaries have no access to modify the training samples and may leads to the backdoor attack being detected easily. In order to explore the backdoor vulnerability of GCNs and create a more practical and stealthy backdoor attack method, this paper proposes a clean-graph backdoor attack against GCNs (CBAG) in the node classification task,which only poisons the training labels without any modification to the training samples, revealing that GCNs have this security vulnerability. Specifically, CBAG designs a new trigger exploration method to find important feature dimensions as the trigger patterns to improve the attack performance. By poisoning the training labels, a hidden backdoor is injected into the GCNs model. Experimental results show that our clean graph backdoor can achieve 99% attack success rate while maintaining the functionality of the GCNs model on benign samples.
2102.04762
Linwei Ye
Linwei Ye, Mrigank Rochan, Zhi Liu, Xiaoqin Zhang and Yang Wang
Referring Segmentation in Images and Videos with Cross-Modal Self-Attention Network
14 pages, 8 figures. arXiv admin note: substantial text overlap with arXiv:1904.04745
null
10.1109/TPAMI.2021.3054384
null
cs.CV
http://creativecommons.org/licenses/by/4.0/
We consider the problem of referring segmentation in images and videos with natural language. Given an input image (or video) and a referring expression, the goal is to segment the entity referred by the expression in the image or video. In this paper, we propose a cross-modal self-attention (CMSA) module to utilize fine details of individual words and the input image or video, which effectively captures the long-range dependencies between linguistic and visual features. Our model can adaptively focus on informative words in the referring expression and important regions in the visual input. We further propose a gated multi-level fusion (GMLF) module to selectively integrate self-attentive cross-modal features corresponding to different levels of visual features. This module controls the feature fusion of information flow of features at different levels with high-level and low-level semantic information related to different attentive words. Besides, we introduce cross-frame self-attention (CFSA) module to effectively integrate temporal information in consecutive frames which extends our method in the case of referring segmentation in videos. Experiments on benchmark datasets of four referring image datasets and two actor and action video segmentation datasets consistently demonstrate that our proposed approach outperforms existing state-of-the-art methods.
[ { "created": "Tue, 9 Feb 2021 11:27:59 GMT", "version": "v1" } ]
2021-02-10
[ [ "Ye", "Linwei", "" ], [ "Rochan", "Mrigank", "" ], [ "Liu", "Zhi", "" ], [ "Zhang", "Xiaoqin", "" ], [ "Wang", "Yang", "" ] ]
We consider the problem of referring segmentation in images and videos with natural language. Given an input image (or video) and a referring expression, the goal is to segment the entity referred by the expression in the image or video. In this paper, we propose a cross-modal self-attention (CMSA) module to utilize fine details of individual words and the input image or video, which effectively captures the long-range dependencies between linguistic and visual features. Our model can adaptively focus on informative words in the referring expression and important regions in the visual input. We further propose a gated multi-level fusion (GMLF) module to selectively integrate self-attentive cross-modal features corresponding to different levels of visual features. This module controls the feature fusion of information flow of features at different levels with high-level and low-level semantic information related to different attentive words. Besides, we introduce cross-frame self-attention (CFSA) module to effectively integrate temporal information in consecutive frames which extends our method in the case of referring segmentation in videos. Experiments on benchmark datasets of four referring image datasets and two actor and action video segmentation datasets consistently demonstrate that our proposed approach outperforms existing state-of-the-art methods.
1112.0958
Qianxue Wang
Jacques M. Bahi, Xiaole Fang, Christophe Guyeux, and Qianxue Wang
On the design of a family of CI pseudo-random number generators
4 pages, In WICOM'11, 7th Int. IEEE Conf. on Wireless Communications, Networking and Mobile Computing, Wuhan, China, pages 1--4, September 2011
null
10.1109/wicom.2011.6040161
null
cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Chaos and its applications in the field of secure communications have attracted a lot of attention. Chaos-based pseudo-random number generators are critical to guarantee security over open networks as the Internet. We have previously demonstrated that it is possible to define such generators with good statistical properties by using a tool called "chaotic iterations", which depends on an iteration function. An approach to find update functions such that the associated generator presents a random-like and chaotic behavior is proposed in this research work. To do so, we use the vectorial Boolean negation as a prototype and explain how to modify this iteration function without deflating the good properties of the associated generator. Simulation results and basic security analysis are then presented to evaluate the randomness of this new family of generators.
[ { "created": "Mon, 5 Dec 2011 15:07:16 GMT", "version": "v1" } ]
2016-11-17
[ [ "Bahi", "Jacques M.", "" ], [ "Fang", "Xiaole", "" ], [ "Guyeux", "Christophe", "" ], [ "Wang", "Qianxue", "" ] ]
Chaos and its applications in the field of secure communications have attracted a lot of attention. Chaos-based pseudo-random number generators are critical to guarantee security over open networks as the Internet. We have previously demonstrated that it is possible to define such generators with good statistical properties by using a tool called "chaotic iterations", which depends on an iteration function. An approach to find update functions such that the associated generator presents a random-like and chaotic behavior is proposed in this research work. To do so, we use the vectorial Boolean negation as a prototype and explain how to modify this iteration function without deflating the good properties of the associated generator. Simulation results and basic security analysis are then presented to evaluate the randomness of this new family of generators.
1604.07044
Xianming Liu
Xianming Liu, Min-Hsuan Tsai, Thomas Huang
Analyzing User Preference for Social Image Recommendation
null
null
null
null
cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the incredibly growing amount of multimedia data shared on the social media platforms, recommender systems have become an important necessity to ease users' burden on the information overload. In such a scenario, extensive amount of heterogeneous information such as tags, image content, in addition to the user-to-item preferences, is extremely valuable for making effective recommendations. In this paper, we explore a novel hybrid algorithm termed {\em STM}, for image recommendation. STM jointly considers the problem of image content analysis with the users' preferences on the basis of sparse representation. STM is able to tackle the challenges of highly sparse user feedbacks and cold-start problmes in the social network scenario. In addition, our model is based on the classical probabilistic matrix factorization and can be easily extended to incorporate other useful information such as the social relationships. We evaluate our approach with a newly collected 0.3 million social image data set from Flickr. The experimental results demonstrate that sparse topic modeling of the image content leads to more effective recommendations, , with a significant performance gain over the state-of-the-art alternatives.
[ { "created": "Sun, 24 Apr 2016 15:54:02 GMT", "version": "v1" } ]
2016-04-26
[ [ "Liu", "Xianming", "" ], [ "Tsai", "Min-Hsuan", "" ], [ "Huang", "Thomas", "" ] ]
With the incredibly growing amount of multimedia data shared on the social media platforms, recommender systems have become an important necessity to ease users' burden on the information overload. In such a scenario, extensive amount of heterogeneous information such as tags, image content, in addition to the user-to-item preferences, is extremely valuable for making effective recommendations. In this paper, we explore a novel hybrid algorithm termed {\em STM}, for image recommendation. STM jointly considers the problem of image content analysis with the users' preferences on the basis of sparse representation. STM is able to tackle the challenges of highly sparse user feedbacks and cold-start problmes in the social network scenario. In addition, our model is based on the classical probabilistic matrix factorization and can be easily extended to incorporate other useful information such as the social relationships. We evaluate our approach with a newly collected 0.3 million social image data set from Flickr. The experimental results demonstrate that sparse topic modeling of the image content leads to more effective recommendations, , with a significant performance gain over the state-of-the-art alternatives.
1504.00037
Alex Horn
Alex Horn and Daniel Kroening
On partial order semantics for SAT/SMT-based symbolic encodings of weak memory concurrency
15 pages, 3 figures
null
null
null
cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Concurrent systems are notoriously difficult to analyze, and technological advances such as weak memory architectures greatly compound this problem. This has renewed interest in partial order semantics as a theoretical foundation for formal verification techniques. Among these, symbolic techniques have been shown to be particularly effective at finding concurrency-related bugs because they can leverage highly optimized decision procedures such as SAT/SMT solvers. This paper gives new fundamental results on partial order semantics for SAT/SMT-based symbolic encodings of weak memory concurrency. In particular, we give the theoretical basis for a decision procedure that can handle a fragment of concurrent programs endowed with least fixed point operators. In addition, we show that a certain partial order semantics of relaxed sequential consistency is equivalent to the conjunction of three extensively studied weak memory axioms by Alglave et al. An important consequence of this equivalence is an asymptotically smaller symbolic encoding for bounded model checking which has only a quadratic number of partial order constraints compared to the state-of-the-art cubic-size encoding.
[ { "created": "Tue, 31 Mar 2015 21:03:30 GMT", "version": "v1" } ]
2015-04-02
[ [ "Horn", "Alex", "" ], [ "Kroening", "Daniel", "" ] ]
Concurrent systems are notoriously difficult to analyze, and technological advances such as weak memory architectures greatly compound this problem. This has renewed interest in partial order semantics as a theoretical foundation for formal verification techniques. Among these, symbolic techniques have been shown to be particularly effective at finding concurrency-related bugs because they can leverage highly optimized decision procedures such as SAT/SMT solvers. This paper gives new fundamental results on partial order semantics for SAT/SMT-based symbolic encodings of weak memory concurrency. In particular, we give the theoretical basis for a decision procedure that can handle a fragment of concurrent programs endowed with least fixed point operators. In addition, we show that a certain partial order semantics of relaxed sequential consistency is equivalent to the conjunction of three extensively studied weak memory axioms by Alglave et al. An important consequence of this equivalence is an asymptotically smaller symbolic encoding for bounded model checking which has only a quadratic number of partial order constraints compared to the state-of-the-art cubic-size encoding.
1801.00377
Walid Shalaby
Walid Shalaby, BahaaEddin AlAila, Mohammed Korayem, Layla Pournajaf, Khalifeh AlJadda, Shannon Quinn, and Wlodek Zadrozny
Help Me Find a Job: A Graph-based Approach for Job Recommendation at Scale
Accepted at 2017 IEEE International Conference on Big Data (BIGDATA)
null
null
null
cs.IR cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Online job boards are one of the central components of modern recruitment industry. With millions of candidates browsing through job postings everyday, the need for accurate, effective, meaningful, and transparent job recommendations is apparent more than ever. While recommendation systems are successfully advancing in variety of online domains by creating social and commercial value, the job recommendation domain is less explored. Existing systems are mostly focused on content analysis of resumes and job descriptions, relying heavily on the accuracy and coverage of the semantic analysis and modeling of the content in which case, they end up usually suffering from rigidity and the lack of implicit semantic relations that are uncovered from users' behavior and could be captured by Collaborative Filtering (CF) methods. Few works which utilize CF do not address the scalability challenges of real-world systems and the problem of cold-start. In this paper, we propose a scalable item-based recommendation system for online job recommendations. Our approach overcomes the major challenges of sparsity and scalability by leveraging a directed graph of jobs connected by multi-edges representing various behavioral and contextual similarity signals. The short lived nature of the items (jobs) in the system and the rapid rate in which new users and jobs enter the system make the cold-start a serious problem hindering CF methods. We address this problem by harnessing the power of deep learning in addition to user behavior to serve hybrid recommendations. Our technique has been leveraged by CareerBuilder.com which is one of the largest job boards in the world to generate high-quality recommendations for millions of users.
[ { "created": "Mon, 1 Jan 2018 00:47:44 GMT", "version": "v1" } ]
2018-01-03
[ [ "Shalaby", "Walid", "" ], [ "AlAila", "BahaaEddin", "" ], [ "Korayem", "Mohammed", "" ], [ "Pournajaf", "Layla", "" ], [ "AlJadda", "Khalifeh", "" ], [ "Quinn", "Shannon", "" ], [ "Zadrozny", "Wlodek", "" ] ]
Online job boards are one of the central components of modern recruitment industry. With millions of candidates browsing through job postings everyday, the need for accurate, effective, meaningful, and transparent job recommendations is apparent more than ever. While recommendation systems are successfully advancing in variety of online domains by creating social and commercial value, the job recommendation domain is less explored. Existing systems are mostly focused on content analysis of resumes and job descriptions, relying heavily on the accuracy and coverage of the semantic analysis and modeling of the content in which case, they end up usually suffering from rigidity and the lack of implicit semantic relations that are uncovered from users' behavior and could be captured by Collaborative Filtering (CF) methods. Few works which utilize CF do not address the scalability challenges of real-world systems and the problem of cold-start. In this paper, we propose a scalable item-based recommendation system for online job recommendations. Our approach overcomes the major challenges of sparsity and scalability by leveraging a directed graph of jobs connected by multi-edges representing various behavioral and contextual similarity signals. The short lived nature of the items (jobs) in the system and the rapid rate in which new users and jobs enter the system make the cold-start a serious problem hindering CF methods. We address this problem by harnessing the power of deep learning in addition to user behavior to serve hybrid recommendations. Our technique has been leveraged by CareerBuilder.com which is one of the largest job boards in the world to generate high-quality recommendations for millions of users.
2201.06801
Subhasis Koley
Susobhan Bandopadhyay, Sasthi C. Ghosh and Subhasis Koley
Improved Bounds on the Span of $L(1,2)$-edge Labeling of Some Infinite Regular Grids
null
null
null
null
cs.DM math.CO
http://creativecommons.org/licenses/by/4.0/
For two given nonnegative integers $h$ and $k$, an $L(h,k)$-edge labeling of a graph $G$ is the assignment of labels $\{0,1, \cdots, n\}$ to the edges so that two edges having a common vertex are labeled with difference at least $h$ and two edges not having any common vertex but having a common edge connecting them are labeled with difference at least $k$. The span $\lambda'_{h,k}{(G)}$ is the minimum $n$ such that $G$ admits an $L(h,k)$-edge labeling. Here our main focus is on finding $\lambda'_{1,2}{(G)}$ for $L(1,2)$-edge labeling of infinite regular hexagonal ($T_3$), square ($T_4$), triangular ($T_6$) and octagonal ($T_8$) grids. It was known that $7 \leq \lambda'_{1,2}{(T_3)} \leq 8$, $10 \leq \lambda'_{1,2}{(T_4)} \leq 11$, $16 \leq \lambda'_{1,2}{(T_6)} \leq 20$ and $25 \leq \lambda'_{1,2}{(T_8)} \leq 28$. Here we settle two long standing open questions i.e. $\lambda'_{1,2}{(T_3)}$ and $\lambda'_{1,2}{(T_4)}$. We show $\lambda'_{1,2}{(T_3)} =7$, $\lambda'_{1,2}{(T_4)}= 11$. We also improve the bound for $T_6$ and $T_8$ and prove $\lambda'_{1,2}{(T_6)} \geq 18$, $ \lambda'_{1,2}{(T_8)} \geq 26$.
[ { "created": "Tue, 18 Jan 2022 07:58:13 GMT", "version": "v1" } ]
2022-01-19
[ [ "Bandopadhyay", "Susobhan", "" ], [ "Ghosh", "Sasthi C.", "" ], [ "Koley", "Subhasis", "" ] ]
For two given nonnegative integers $h$ and $k$, an $L(h,k)$-edge labeling of a graph $G$ is the assignment of labels $\{0,1, \cdots, n\}$ to the edges so that two edges having a common vertex are labeled with difference at least $h$ and two edges not having any common vertex but having a common edge connecting them are labeled with difference at least $k$. The span $\lambda'_{h,k}{(G)}$ is the minimum $n$ such that $G$ admits an $L(h,k)$-edge labeling. Here our main focus is on finding $\lambda'_{1,2}{(G)}$ for $L(1,2)$-edge labeling of infinite regular hexagonal ($T_3$), square ($T_4$), triangular ($T_6$) and octagonal ($T_8$) grids. It was known that $7 \leq \lambda'_{1,2}{(T_3)} \leq 8$, $10 \leq \lambda'_{1,2}{(T_4)} \leq 11$, $16 \leq \lambda'_{1,2}{(T_6)} \leq 20$ and $25 \leq \lambda'_{1,2}{(T_8)} \leq 28$. Here we settle two long standing open questions i.e. $\lambda'_{1,2}{(T_3)}$ and $\lambda'_{1,2}{(T_4)}$. We show $\lambda'_{1,2}{(T_3)} =7$, $\lambda'_{1,2}{(T_4)}= 11$. We also improve the bound for $T_6$ and $T_8$ and prove $\lambda'_{1,2}{(T_6)} \geq 18$, $ \lambda'_{1,2}{(T_8)} \geq 26$.
1908.10042
EPTCS
Wolfgang Ahrendt (Chalmers University of Technology), Ludovic Henrio (Univ Lyon, EnsL, UCBL, CNRS, Inria, LIP), Wytse Oortwijn (University of Twente)
Who is to Blame? Runtime Verification of Distributed Objects with Active Monitors
In Proceedings VORTEX 2018, arXiv:1908.09302
EPTCS 302, 2019, pp. 32-46
10.4204/EPTCS.302.3
null
cs.SE cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Since distributed software systems are ubiquitous, their correct functioning is crucially important. Static verification is possible in principle, but requires high expertise and effort which is not feasible in many eco-systems. Runtime verification can serve as a lean alternative, where monitoring mechanisms are automatically generated from property specifications, to check compliance at runtime. This paper contributes a practical solution for powerful and flexible runtime verification of distributed, object-oriented applications, via a combination of the runtime verification tool Larva and the active object framework ProActive. Even if Larva supports in itself only the generation of local, sequential monitors, we empower Larva for distributed monitoring by connecting monitors with active objects, turning them into active, communicating monitors. We discuss how this allows for a variety of monitoring architectures. Further, we show how property specifications, and thereby the generated monitors, provide a model that splits the blame between the local object and its environment. While Larva itself focuses on monitoring of control-oriented properties, we use the Larva front-end StaRVOOrS to also capture data-oriented (pre/post) properties in the distributed monitoring. We demonstrate this approach to distributed runtime verification with a case study, a distributed key/value store.
[ { "created": "Tue, 27 Aug 2019 06:20:22 GMT", "version": "v1" } ]
2019-08-28
[ [ "Ahrendt", "Wolfgang", "", "Chalmers University of Technology" ], [ "Henrio", "Ludovic", "", "Univ Lyon, EnsL, UCBL, CNRS, Inria, LIP" ], [ "Oortwijn", "Wytse", "", "University of\n Twente" ] ]
Since distributed software systems are ubiquitous, their correct functioning is crucially important. Static verification is possible in principle, but requires high expertise and effort which is not feasible in many eco-systems. Runtime verification can serve as a lean alternative, where monitoring mechanisms are automatically generated from property specifications, to check compliance at runtime. This paper contributes a practical solution for powerful and flexible runtime verification of distributed, object-oriented applications, via a combination of the runtime verification tool Larva and the active object framework ProActive. Even if Larva supports in itself only the generation of local, sequential monitors, we empower Larva for distributed monitoring by connecting monitors with active objects, turning them into active, communicating monitors. We discuss how this allows for a variety of monitoring architectures. Further, we show how property specifications, and thereby the generated monitors, provide a model that splits the blame between the local object and its environment. While Larva itself focuses on monitoring of control-oriented properties, we use the Larva front-end StaRVOOrS to also capture data-oriented (pre/post) properties in the distributed monitoring. We demonstrate this approach to distributed runtime verification with a case study, a distributed key/value store.
1911.02590
Jonathan Lorraine
Jonathan Lorraine, Paul Vicol, David Duvenaud
Optimizing Millions of Hyperparameters by Implicit Differentiation
Submitted to AISTATS 2020
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose an algorithm for inexpensive gradient-based hyperparameter optimization that combines the implicit function theorem (IFT) with efficient inverse Hessian approximations. We present results about the relationship between the IFT and differentiating through optimization, motivating our algorithm. We use the proposed approach to train modern network architectures with millions of weights and millions of hyper-parameters. For example, we learn a data-augmentation network - where every weight is a hyperparameter tuned for validation performance - outputting augmented training examples. Jointly tuning weights and hyperparameters with our approach is only a few times more costly in memory and compute than standard training.
[ { "created": "Wed, 6 Nov 2019 19:04:16 GMT", "version": "v1" } ]
2019-11-11
[ [ "Lorraine", "Jonathan", "" ], [ "Vicol", "Paul", "" ], [ "Duvenaud", "David", "" ] ]
We propose an algorithm for inexpensive gradient-based hyperparameter optimization that combines the implicit function theorem (IFT) with efficient inverse Hessian approximations. We present results about the relationship between the IFT and differentiating through optimization, motivating our algorithm. We use the proposed approach to train modern network architectures with millions of weights and millions of hyper-parameters. For example, we learn a data-augmentation network - where every weight is a hyperparameter tuned for validation performance - outputting augmented training examples. Jointly tuning weights and hyperparameters with our approach is only a few times more costly in memory and compute than standard training.
2012.12507
Se Young Chun
Dongwon Park, Dong Un Kang, Se Young Chun
Blur More To Deblur Better: Multi-Blur2Deblur For Efficient Video Deblurring
9 pages, 7 figures
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
One of the key components for video deblurring is how to exploit neighboring frames. Recent state-of-the-art methods either used aligned adjacent frames to the center frame or propagated the information on past frames to the current frame recurrently. Here we propose multi-blur-to-deblur (MB2D), a novel concept to exploit neighboring frames for efficient video deblurring. Firstly, inspired by unsharp masking, we argue that using more blurred images with long exposures as additional inputs significantly improves performance. Secondly, we propose multi-blurring recurrent neural network (MBRNN) that can synthesize more blurred images from neighboring frames, yielding substantially improved performance with existing video deblurring methods. Lastly, we propose multi-scale deblurring with connecting recurrent feature map from MBRNN (MSDR) to achieve state-of-the-art performance on the popular GoPro and Su datasets in fast and memory efficient ways.
[ { "created": "Wed, 23 Dec 2020 06:17:31 GMT", "version": "v1" } ]
2020-12-24
[ [ "Park", "Dongwon", "" ], [ "Kang", "Dong Un", "" ], [ "Chun", "Se Young", "" ] ]
One of the key components for video deblurring is how to exploit neighboring frames. Recent state-of-the-art methods either used aligned adjacent frames to the center frame or propagated the information on past frames to the current frame recurrently. Here we propose multi-blur-to-deblur (MB2D), a novel concept to exploit neighboring frames for efficient video deblurring. Firstly, inspired by unsharp masking, we argue that using more blurred images with long exposures as additional inputs significantly improves performance. Secondly, we propose multi-blurring recurrent neural network (MBRNN) that can synthesize more blurred images from neighboring frames, yielding substantially improved performance with existing video deblurring methods. Lastly, we propose multi-scale deblurring with connecting recurrent feature map from MBRNN (MSDR) to achieve state-of-the-art performance on the popular GoPro and Su datasets in fast and memory efficient ways.
2307.14686
Josep Marti-Saumell
Josep Mart\'i-Saumell, Hugo Duarte, Patrick Grosch, Juan Andrade-Cetto, Angel Santamaria-Navarro, Joan Sol\`a
Borinot: an open thrust-torque-controlled robot for research on agile aerial-contact motion
14 pages, 13 figures. See related video at https://youtu.be/Ob7IIVB6P_A
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper introduces Borinot, an open-source aerial robotic platform designed to conduct research on hybrid agile locomotion and manipulation using flight and contacts. This platform features an agile and powerful hexarotor that can be outfitted with torque-actuated limbs of diverse architecture, allowing for whole-body dynamic control. As a result, Borinot can perform agile tasks such as aggressive or acrobatic maneuvers with the participation of the whole-body dynamics. The limbs attached to Borinot can be utilized in various ways; during contact, they can be used as legs to create contact-based locomotion, or as arms to manipulate objects. In free flight, they can be used as tails to contribute to dynamics, mimicking the movements of many animals. This allows for any hybridization of these dynamic modes, making Borinot an ideal open-source platform for research on hybrid aerial-contact agile motion. To demonstrate the key capabilities of Borinot in terms of agility with hybrid motion modes, we have fitted a planar 2DoF limb and implemented a whole-body torque-level model-predictive-control. The result is a capable and adaptable platform that, we believe, opens up new avenues of research in the field of agile robotics. Interesting links\footnote{Documentation: \url{www.iri.upc.edu/borinot}}\footnote{Video: \url{https://youtu.be/Ob7IIVB6P_A}}.
[ { "created": "Thu, 27 Jul 2023 08:19:47 GMT", "version": "v1" } ]
2023-07-28
[ [ "Martí-Saumell", "Josep", "" ], [ "Duarte", "Hugo", "" ], [ "Grosch", "Patrick", "" ], [ "Andrade-Cetto", "Juan", "" ], [ "Santamaria-Navarro", "Angel", "" ], [ "Solà", "Joan", "" ] ]
This paper introduces Borinot, an open-source aerial robotic platform designed to conduct research on hybrid agile locomotion and manipulation using flight and contacts. This platform features an agile and powerful hexarotor that can be outfitted with torque-actuated limbs of diverse architecture, allowing for whole-body dynamic control. As a result, Borinot can perform agile tasks such as aggressive or acrobatic maneuvers with the participation of the whole-body dynamics. The limbs attached to Borinot can be utilized in various ways; during contact, they can be used as legs to create contact-based locomotion, or as arms to manipulate objects. In free flight, they can be used as tails to contribute to dynamics, mimicking the movements of many animals. This allows for any hybridization of these dynamic modes, making Borinot an ideal open-source platform for research on hybrid aerial-contact agile motion. To demonstrate the key capabilities of Borinot in terms of agility with hybrid motion modes, we have fitted a planar 2DoF limb and implemented a whole-body torque-level model-predictive-control. The result is a capable and adaptable platform that, we believe, opens up new avenues of research in the field of agile robotics. Interesting links\footnote{Documentation: \url{www.iri.upc.edu/borinot}}\footnote{Video: \url{https://youtu.be/Ob7IIVB6P_A}}.
2208.11483
Suncheng Xiang
Hongwei Xu, Suncheng Xiang, Dahong Qian
SubFace: Learning with Softmax Approximation for Face Recognition
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The softmax-based loss functions and its variants (e.g., cosface, sphereface, and arcface) significantly improve the face recognition performance in wild unconstrained scenes. A common practice of these algorithms is to perform optimizations on the multiplication between the embedding features and the linear transformation matrix. However in most cases, the dimension of embedding features is given based on traditional design experience, and there is less-studied on improving performance using the feature itself when giving a fixed size. To address this challenge, this paper presents a softmax approximation method called SubFace, which employs the subspace feature to promote the performance of face recognition. Specifically, we dynamically select the non-overlapping subspace features in each batch during training, and then use the subspace features to approximate full-feature among softmax-based loss, so the discriminability of the deep model can be significantly enhanced for face recognition. Comprehensive experiments conducted on benchmark datasets demonstrate that our method can significantly improve the performance of vanilla CNN baseline, which strongly proves the effectiveness of subspace strategy with the margin-based loss.
[ { "created": "Wed, 24 Aug 2022 12:31:08 GMT", "version": "v1" } ]
2022-08-25
[ [ "Xu", "Hongwei", "" ], [ "Xiang", "Suncheng", "" ], [ "Qian", "Dahong", "" ] ]
The softmax-based loss functions and its variants (e.g., cosface, sphereface, and arcface) significantly improve the face recognition performance in wild unconstrained scenes. A common practice of these algorithms is to perform optimizations on the multiplication between the embedding features and the linear transformation matrix. However in most cases, the dimension of embedding features is given based on traditional design experience, and there is less-studied on improving performance using the feature itself when giving a fixed size. To address this challenge, this paper presents a softmax approximation method called SubFace, which employs the subspace feature to promote the performance of face recognition. Specifically, we dynamically select the non-overlapping subspace features in each batch during training, and then use the subspace features to approximate full-feature among softmax-based loss, so the discriminability of the deep model can be significantly enhanced for face recognition. Comprehensive experiments conducted on benchmark datasets demonstrate that our method can significantly improve the performance of vanilla CNN baseline, which strongly proves the effectiveness of subspace strategy with the margin-based loss.
1711.02950
Federico Costantini
Federico Costantini
MaaS and GDPR: an overview
Paper (10 pages) accepted at the international conference Intelligent Transport Systems. From research and development to the market uptake, Helsinki (Finland) November 29/30, 2017. Not published nor in press
null
null
null
cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In MaaS (Mobility as a Service), means of transport are virtualized in mobility resources and provided to users using the Internet. From a legal perspective, this model of ITS (Intelligent Transport System) raises several concerns with regard to data protection. This contribution, after a short description of MaaS and an introduction to the issues of data protection in ITS, explores the impact of GDPR (General Data Protection Regulation) in the European Union, detecting possible threats and remedies and suggesting a plausible approach.
[ { "created": "Wed, 8 Nov 2017 14:14:15 GMT", "version": "v1" } ]
2017-11-09
[ [ "Costantini", "Federico", "" ] ]
In MaaS (Mobility as a Service), means of transport are virtualized in mobility resources and provided to users using the Internet. From a legal perspective, this model of ITS (Intelligent Transport System) raises several concerns with regard to data protection. This contribution, after a short description of MaaS and an introduction to the issues of data protection in ITS, explores the impact of GDPR (General Data Protection Regulation) in the European Union, detecting possible threats and remedies and suggesting a plausible approach.
2104.06772
Anthony Ngo
Anthony Ngo, Max Paul Bauer, Michael Resch
Deep Evaluation Metric: Learning to Evaluate Simulated Radar Point Clouds for Virtual Testing of Autonomous Driving
2021 IEEE Radar Conference (IEEE RadarConf 2021)
null
10.1109/RadarConf2147009.2021.9455235
null
cs.CV cs.AI eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The usage of environment sensor models for virtual testing is a promising approach to reduce the testing effort of autonomous driving. However, in order to deduce any statements regarding the performance of an autonomous driving function based on simulation, the sensor model has to be validated to determine the discrepancy between the synthetic and real sensor data. Since a certain degree of divergence can be assumed to exist, the sufficient level of fidelity must be determined, which poses a major challenge. In particular, a method for quantifying the fidelity of a sensor model does not exist and the problem of defining an appropriate metric remains. In this work, we train a neural network to distinguish real and simulated radar sensor data with the purpose of learning the latent features of real radar point clouds. Furthermore, we propose the classifier's confidence score for the `real radar point cloud' class as a metric to determine the degree of fidelity of synthetically generated radar data. The presented approach is evaluated and it can be demonstrated that the proposed deep evaluation metric outperforms conventional metrics in terms of its capability to identify characteristic differences between real and simulated radar data.
[ { "created": "Wed, 14 Apr 2021 11:04:50 GMT", "version": "v1" }, { "created": "Mon, 21 Jun 2021 08:30:09 GMT", "version": "v2" } ]
2021-06-22
[ [ "Ngo", "Anthony", "" ], [ "Bauer", "Max Paul", "" ], [ "Resch", "Michael", "" ] ]
The usage of environment sensor models for virtual testing is a promising approach to reduce the testing effort of autonomous driving. However, in order to deduce any statements regarding the performance of an autonomous driving function based on simulation, the sensor model has to be validated to determine the discrepancy between the synthetic and real sensor data. Since a certain degree of divergence can be assumed to exist, the sufficient level of fidelity must be determined, which poses a major challenge. In particular, a method for quantifying the fidelity of a sensor model does not exist and the problem of defining an appropriate metric remains. In this work, we train a neural network to distinguish real and simulated radar sensor data with the purpose of learning the latent features of real radar point clouds. Furthermore, we propose the classifier's confidence score for the `real radar point cloud' class as a metric to determine the degree of fidelity of synthetically generated radar data. The presented approach is evaluated and it can be demonstrated that the proposed deep evaluation metric outperforms conventional metrics in terms of its capability to identify characteristic differences between real and simulated radar data.
2311.10529
Yichi Zhang
Yichi Zhang, Shiyao Hu, Sijie Ren, Chen Jiang, Yuan Cheng, Yuan Qi
Enhancing the Reliability of Segment Anything Model for Auto-Prompting Medical Image Segmentation with Uncertainty Rectification
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Segment Anything Model (SAM) has recently emerged as a groundbreaking foundation model for prompt-driven image segmentation tasks. However, both the original SAM and its medical variants require slice-by-slice manual prompting of target structures, which directly increase the burden for applications. Despite attempts of auto-prompting to turn SAM into a fully automatic manner, it still exhibits subpar performance and lacks of reliability especially in the field of medical imaging. In this paper, we propose UR-SAM, an uncertainty rectified SAM framework to enhance the reliability for auto-prompting medical image segmentation. Building upon a localization framework for automatic prompt generation, our method incorporates a prompt augmentation module to obtain a series of input prompts for SAM for uncertainty estimation and an uncertainty-based rectification module to further utilize the distribution of estimated uncertainty to improve the segmentation performance. Extensive experiments on two public 3D medical datasets covering the segmentation of 35 organs demonstrate that without supplementary training or fine-tuning, our method further improves the segmentation performance with up to 10.7 % and 13.8 % in dice similarity coefficient, demonstrating efficiency and broad capabilities for medical image segmentation without manual prompting.
[ { "created": "Fri, 17 Nov 2023 13:49:00 GMT", "version": "v1" }, { "created": "Wed, 13 Dec 2023 04:57:47 GMT", "version": "v2" }, { "created": "Mon, 18 Mar 2024 08:47:03 GMT", "version": "v3" } ]
2024-03-19
[ [ "Zhang", "Yichi", "" ], [ "Hu", "Shiyao", "" ], [ "Ren", "Sijie", "" ], [ "Jiang", "Chen", "" ], [ "Cheng", "Yuan", "" ], [ "Qi", "Yuan", "" ] ]
The Segment Anything Model (SAM) has recently emerged as a groundbreaking foundation model for prompt-driven image segmentation tasks. However, both the original SAM and its medical variants require slice-by-slice manual prompting of target structures, which directly increase the burden for applications. Despite attempts of auto-prompting to turn SAM into a fully automatic manner, it still exhibits subpar performance and lacks of reliability especially in the field of medical imaging. In this paper, we propose UR-SAM, an uncertainty rectified SAM framework to enhance the reliability for auto-prompting medical image segmentation. Building upon a localization framework for automatic prompt generation, our method incorporates a prompt augmentation module to obtain a series of input prompts for SAM for uncertainty estimation and an uncertainty-based rectification module to further utilize the distribution of estimated uncertainty to improve the segmentation performance. Extensive experiments on two public 3D medical datasets covering the segmentation of 35 organs demonstrate that without supplementary training or fine-tuning, our method further improves the segmentation performance with up to 10.7 % and 13.8 % in dice similarity coefficient, demonstrating efficiency and broad capabilities for medical image segmentation without manual prompting.
1809.09849
Richard Torkar
Richard Torkar, Carlo A. Furia, Robert Feldt, Francisco Gomes de Oliveira Neto, Lucas Gren, Per Lenberg, Neil A. Ernst
A Method to Assess and Argue for Practical Significance in Software Engineering
13 pages, 9 figures, 3 tables. Minor rev update
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A key goal of empirical research in software engineering is to assess practical significance, which answers whether the observed effects of some compared treatments show a relevant difference in practice in realistic scenarios. Even though plenty of standard techniques exist to assess statistical significance, connecting it to practical significance is not straightforward or routinely done; indeed, only a few empirical studies in software engineering assess practical significance in a principled and systematic way. In this paper, we argue that Bayesian data analysis provides suitable tools to assess practical significance rigorously. We demonstrate our claims in a case study comparing different test techniques. The case study's data was previously analyzed (Afzal et al., 2015) using standard techniques focusing on statistical significance. Here, we build a multilevel model of the same data, which we fit and validate using Bayesian techniques. Our method is to apply cumulative prospect theory on top of the statistical model to quantitatively connect our statistical analysis output to a practically meaningful context. This is then the basis both for assessing and arguing for practical significance. Our study demonstrates that Bayesian analysis provides a technically rigorous yet practical framework for empirical software engineering. A substantial side effect is that any uncertainty in the underlying data will be propagated through the statistical model, and its effects on practical significance are made clear. Thus, in combination with cumulative prospect theory, Bayesian analysis supports seamlessly assessing practical significance in an empirical software engineering context, thus potentially clarifying and extending the relevance of research for practitioners.
[ { "created": "Wed, 26 Sep 2018 08:39:46 GMT", "version": "v1" }, { "created": "Wed, 18 Mar 2020 10:35:12 GMT", "version": "v2" }, { "created": "Fri, 3 Apr 2020 14:50:33 GMT", "version": "v3" }, { "created": "Wed, 29 Apr 2020 14:11:03 GMT", "version": "v4" }, { "created": "Sat, 31 Oct 2020 08:44:00 GMT", "version": "v5" }, { "created": "Tue, 3 Nov 2020 08:57:13 GMT", "version": "v6" }, { "created": "Fri, 25 Dec 2020 13:35:33 GMT", "version": "v7" } ]
2020-12-29
[ [ "Torkar", "Richard", "" ], [ "Furia", "Carlo A.", "" ], [ "Feldt", "Robert", "" ], [ "Neto", "Francisco Gomes de Oliveira", "" ], [ "Gren", "Lucas", "" ], [ "Lenberg", "Per", "" ], [ "Ernst", "Neil A.", "" ] ]
A key goal of empirical research in software engineering is to assess practical significance, which answers whether the observed effects of some compared treatments show a relevant difference in practice in realistic scenarios. Even though plenty of standard techniques exist to assess statistical significance, connecting it to practical significance is not straightforward or routinely done; indeed, only a few empirical studies in software engineering assess practical significance in a principled and systematic way. In this paper, we argue that Bayesian data analysis provides suitable tools to assess practical significance rigorously. We demonstrate our claims in a case study comparing different test techniques. The case study's data was previously analyzed (Afzal et al., 2015) using standard techniques focusing on statistical significance. Here, we build a multilevel model of the same data, which we fit and validate using Bayesian techniques. Our method is to apply cumulative prospect theory on top of the statistical model to quantitatively connect our statistical analysis output to a practically meaningful context. This is then the basis both for assessing and arguing for practical significance. Our study demonstrates that Bayesian analysis provides a technically rigorous yet practical framework for empirical software engineering. A substantial side effect is that any uncertainty in the underlying data will be propagated through the statistical model, and its effects on practical significance are made clear. Thus, in combination with cumulative prospect theory, Bayesian analysis supports seamlessly assessing practical significance in an empirical software engineering context, thus potentially clarifying and extending the relevance of research for practitioners.
1808.01280
ShihChung Lo Ph.D.
ShihChung B. Lo, Ph.D., Matthew T. Freedman, M.D., Seong K. Mun, Ph.D., and Heang-Ping Chan, Ph.D
Geared Rotationally Identical and Invariant Convolutional Neural Network Systems
14 pages, 6 figures, 8 tables
null
null
null
cs.NE cs.LG cs.SI stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Theorems and techniques to form different types of transformationally invariant processing and to produce the same output quantitatively based on either transformationally invariant operators or symmetric operations have recently been introduced by the authors. In this study, we further propose to compose a geared rotationally identical CNN system (GRI-CNN) with a small step angle by connecting networks of participated processes at the first flatten layer. Using an ordinary CNN structure as a base, requirements for constructing a GRI-CNN include the use of either symmetric input vector or kernels with an angle increment that can form a complete cycle as a "gearwheel". Four basic GRI-CNN structures were studied. Each of them can produce quantitatively identical output results when a rotation angle of the input vector is evenly divisible by the step angle of the gear. Our study showed when an input vector rotated with an angle does not match to a step angle, the GRI-CNN can also produce a highly consistent result. With a design of using an ultra-fine gear-tooth step angle (e.g., 1 degree or 0.1 degree), all four GRI-CNN systems can be constructed virtually isotropically.
[ { "created": "Fri, 3 Aug 2018 02:27:40 GMT", "version": "v1" }, { "created": "Wed, 8 Aug 2018 15:08:37 GMT", "version": "v2" }, { "created": "Fri, 10 Aug 2018 11:26:09 GMT", "version": "v3" } ]
2018-08-13
[ [ "Lo", "ShihChung B.", "" ], [ "D.", "Ph.", "" ], [ "Freedman", "Matthew T.", "" ], [ "D.", "M.", "" ], [ "Mun", "Seong K.", "" ], [ "D.", "Ph.", "" ], [ "Chan", "Heang-Ping", "" ], [ "D", "Ph.", "" ] ]
Theorems and techniques to form different types of transformationally invariant processing and to produce the same output quantitatively based on either transformationally invariant operators or symmetric operations have recently been introduced by the authors. In this study, we further propose to compose a geared rotationally identical CNN system (GRI-CNN) with a small step angle by connecting networks of participated processes at the first flatten layer. Using an ordinary CNN structure as a base, requirements for constructing a GRI-CNN include the use of either symmetric input vector or kernels with an angle increment that can form a complete cycle as a "gearwheel". Four basic GRI-CNN structures were studied. Each of them can produce quantitatively identical output results when a rotation angle of the input vector is evenly divisible by the step angle of the gear. Our study showed when an input vector rotated with an angle does not match to a step angle, the GRI-CNN can also produce a highly consistent result. With a design of using an ultra-fine gear-tooth step angle (e.g., 1 degree or 0.1 degree), all four GRI-CNN systems can be constructed virtually isotropically.
1607.05408
Matthias Gall\'e
Will Radford, Matthias Galle
Discriminating between similar languages in Twitter using label propagation
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Identifying the language of social media messages is an important first step in linguistic processing. Existing models for Twitter focus on content analysis, which is successful for dissimilar language pairs. We propose a label propagation approach that takes the social graph of tweet authors into account as well as content to better tease apart similar languages. This results in state-of-the-art shared task performance of $76.63\%$, $1.4\%$ higher than the top system.
[ { "created": "Tue, 19 Jul 2016 05:38:58 GMT", "version": "v1" } ]
2016-07-20
[ [ "Radford", "Will", "" ], [ "Galle", "Matthias", "" ] ]
Identifying the language of social media messages is an important first step in linguistic processing. Existing models for Twitter focus on content analysis, which is successful for dissimilar language pairs. We propose a label propagation approach that takes the social graph of tweet authors into account as well as content to better tease apart similar languages. This results in state-of-the-art shared task performance of $76.63\%$, $1.4\%$ higher than the top system.
2209.02101
Wolfgang Mulzer
Michaela Borzechowski and Wolfgang Mulzer
Unique Sink Orientations of Grids is in Unique End of Potential Line
8 pages, 5 figures
null
null
null
cs.CG
http://creativecommons.org/licenses/by/4.0/
The complexity classes Unique End of Potential Line (UEOPL) and its promise version PUEOPL were introduced in 2018 by Fearnly et al. UEOPL captures search problems where the instances are promised to have a unique solution. UEOPL captures total search versions of these promise problems. The promise problems can be made total by defining violations that are returned as a short certificate of an unfulfilled promise. GridUSO is the problem of finding the sink in a grid with a unique sink orientation. It was introduced by G\"artner et al. We describe a promise preserving reduction from GridUSO to UniqueForwardEOPL, a UEOPL-complete problem. Thus, we show that GridUSO is in UEOPL and its promise version is in PUEOPL.
[ { "created": "Mon, 5 Sep 2022 19:06:53 GMT", "version": "v1" } ]
2022-09-07
[ [ "Borzechowski", "Michaela", "" ], [ "Mulzer", "Wolfgang", "" ] ]
The complexity classes Unique End of Potential Line (UEOPL) and its promise version PUEOPL were introduced in 2018 by Fearnly et al. UEOPL captures search problems where the instances are promised to have a unique solution. UEOPL captures total search versions of these promise problems. The promise problems can be made total by defining violations that are returned as a short certificate of an unfulfilled promise. GridUSO is the problem of finding the sink in a grid with a unique sink orientation. It was introduced by G\"artner et al. We describe a promise preserving reduction from GridUSO to UniqueForwardEOPL, a UEOPL-complete problem. Thus, we show that GridUSO is in UEOPL and its promise version is in PUEOPL.
0708.1411
Sajad Sadough
Sajad Sadough (LSS), Pablo Piantanida (LSS), Pierre Duhamel (LSS)
Achievable Outage Rates with Improved Decoding of Bicm Multiband Ofdm Under Channel Estimation Errors
null
Dans 40th Asilomar Conference on Signals, Systems, and Computers - 40th Asilomar Conference on Signals, Systems, and Computers, Monterey : \'Etats-Unis d'Am\'erique (2007)
null
null
cs.NI
null
We consider the decoding of bit interleaved coded modulation (BICM) applied to multiband OFDM for practical scenarios where only a noisy (possibly very bad) estimate of the channel is available at the receiver. First, a decoding metric based on the channel it a posteriori probability density, conditioned on the channel estimate is derived and used for decoding BICM multiband OFDM. Then, we characterize the limits of reliable information rates in terms of the maximal achievable outage rates associated to the proposed metric. We also compare our results with the outage rates of a system using a theoretical decoder. Our results are useful for designing a communication system where a prescribed quality of service (QoS), in terms of achievable target rates with small error probability, must be satisfied even in the presence of imperfect channel estimation. Numerical results over both realistic UWB and theoretical Rayleigh fading channels show that the proposed method provides significant gain in terms of BER and outage rates compared to the classical mismatched detector, without introducing any additional complexity.
[ { "created": "Fri, 10 Aug 2007 12:13:51 GMT", "version": "v1" } ]
2007-08-13
[ [ "Sadough", "Sajad", "", "LSS" ], [ "Piantanida", "Pablo", "", "LSS" ], [ "Duhamel", "Pierre", "", "LSS" ] ]
We consider the decoding of bit interleaved coded modulation (BICM) applied to multiband OFDM for practical scenarios where only a noisy (possibly very bad) estimate of the channel is available at the receiver. First, a decoding metric based on the channel it a posteriori probability density, conditioned on the channel estimate is derived and used for decoding BICM multiband OFDM. Then, we characterize the limits of reliable information rates in terms of the maximal achievable outage rates associated to the proposed metric. We also compare our results with the outage rates of a system using a theoretical decoder. Our results are useful for designing a communication system where a prescribed quality of service (QoS), in terms of achievable target rates with small error probability, must be satisfied even in the presence of imperfect channel estimation. Numerical results over both realistic UWB and theoretical Rayleigh fading channels show that the proposed method provides significant gain in terms of BER and outage rates compared to the classical mismatched detector, without introducing any additional complexity.
2110.00785
Delano Oliveira
Delano Oliveira, Reydne Bruno, Fernanda Madeiral, Fernando Castor
Evaluating Code Readability and Legibility: An Examination of Human-centric Studies
null
null
10.1109/ICSME46990.2020.00041
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Reading code is an essential activity in software maintenance and evolution. Several studies with human subjects have investigated how different factors, such as the employed programming constructs and naming conventions, can impact code readability, i.e., what makes a program easier or harder to read and apprehend by developers, and code legibility, i.e., what influences the ease of identifying elements of a program. These studies evaluate readability and legibility by means of different comprehension tasks and response variables. In this paper, we examine these tasks and variables in studies that compare programming constructs, coding idioms, naming conventions, and formatting guidelines, e.g., recursive vs. iterative code. To that end, we have conducted a systematic literature review where we found 54 relevant papers. Most of these studies evaluate code readability and legibility by measuring the correctness of the subjects' results (83.3%) or simply asking their opinions (55.6%). Some studies (16.7%) rely exclusively on the latter variable.There are still few studies that monitor subjects' physical signs, such as brain activation regions (5%). Moreover, our study shows that some variables are multi-faceted. For instance, correctness can be measured as the ability to predict the output of a program, answer questions about its behavior, or recall parts of it. These results make it clear that different evaluation approaches require different competencies from subjects, e.g., tracing the program vs. summarizing its goal vs. memorizing its text. To assist researchers in the design of new studies and improve our comprehension of existing ones, we model program comprehension as a learning activity by adapting a preexisting learning taxonomy. This adaptation indicates that some competencies are often exercised in these evaluations whereas others are rarely targeted.
[ { "created": "Sat, 2 Oct 2021 11:21:15 GMT", "version": "v1" } ]
2021-10-05
[ [ "Oliveira", "Delano", "" ], [ "Bruno", "Reydne", "" ], [ "Madeiral", "Fernanda", "" ], [ "Castor", "Fernando", "" ] ]
Reading code is an essential activity in software maintenance and evolution. Several studies with human subjects have investigated how different factors, such as the employed programming constructs and naming conventions, can impact code readability, i.e., what makes a program easier or harder to read and apprehend by developers, and code legibility, i.e., what influences the ease of identifying elements of a program. These studies evaluate readability and legibility by means of different comprehension tasks and response variables. In this paper, we examine these tasks and variables in studies that compare programming constructs, coding idioms, naming conventions, and formatting guidelines, e.g., recursive vs. iterative code. To that end, we have conducted a systematic literature review where we found 54 relevant papers. Most of these studies evaluate code readability and legibility by measuring the correctness of the subjects' results (83.3%) or simply asking their opinions (55.6%). Some studies (16.7%) rely exclusively on the latter variable.There are still few studies that monitor subjects' physical signs, such as brain activation regions (5%). Moreover, our study shows that some variables are multi-faceted. For instance, correctness can be measured as the ability to predict the output of a program, answer questions about its behavior, or recall parts of it. These results make it clear that different evaluation approaches require different competencies from subjects, e.g., tracing the program vs. summarizing its goal vs. memorizing its text. To assist researchers in the design of new studies and improve our comprehension of existing ones, we model program comprehension as a learning activity by adapting a preexisting learning taxonomy. This adaptation indicates that some competencies are often exercised in these evaluations whereas others are rarely targeted.
2103.10631
Eric Schwenker
Eric Schwenker, Weixin Jiang, Trevor Spreadbury, Nicola Ferrier, Oliver Cossairt, Maria K. Y. Chan
EXSCLAIM! -- An automated pipeline for the construction of labeled materials imaging datasets from literature
null
null
null
null
cs.IR cond-mat.mtrl-sci
http://creativecommons.org/licenses/by/4.0/
Due to recent improvements in image resolution and acquisition speed, materials microscopy is experiencing an explosion of published imaging data. The standard publication format, while sufficient for traditional data ingestion scenarios where a select number of images can be critically examined and curated manually, is not conducive to large-scale data aggregation or analysis, hindering data sharing and reuse. Most images in publications are presented as components of a larger figure with their explicit context buried in the main body or caption text, so even if aggregated, collections of images with weak or no digitized contextual labels have limited value. To solve the problem of curating labeled microscopy data from literature, this work introduces the EXSCLAIM! Python toolkit for the automatic EXtraction, Separation, and Caption-based natural Language Annotation of IMages from scientific literature. We highlight the methodology behind the construction of EXSCLAIM! and demonstrate its ability to extract and label open-source scientific images at high volume.
[ { "created": "Fri, 19 Mar 2021 04:48:12 GMT", "version": "v1" } ]
2021-03-22
[ [ "Schwenker", "Eric", "" ], [ "Jiang", "Weixin", "" ], [ "Spreadbury", "Trevor", "" ], [ "Ferrier", "Nicola", "" ], [ "Cossairt", "Oliver", "" ], [ "Chan", "Maria K. Y.", "" ] ]
Due to recent improvements in image resolution and acquisition speed, materials microscopy is experiencing an explosion of published imaging data. The standard publication format, while sufficient for traditional data ingestion scenarios where a select number of images can be critically examined and curated manually, is not conducive to large-scale data aggregation or analysis, hindering data sharing and reuse. Most images in publications are presented as components of a larger figure with their explicit context buried in the main body or caption text, so even if aggregated, collections of images with weak or no digitized contextual labels have limited value. To solve the problem of curating labeled microscopy data from literature, this work introduces the EXSCLAIM! Python toolkit for the automatic EXtraction, Separation, and Caption-based natural Language Annotation of IMages from scientific literature. We highlight the methodology behind the construction of EXSCLAIM! and demonstrate its ability to extract and label open-source scientific images at high volume.
2306.15951
Zhiyi Zhang
Zhiyi Zhang, Pengfei Zhang, Zhuopin Xu, Qi Wang
Reduce Computational Complexity for Convolutional Layers by Skipping Zeros
null
null
null
null
cs.LG cs.AI
http://creativecommons.org/licenses/by-nc-nd/4.0/
Convolutional neural networks necessitate good algorithms to reduce complexity, and sufficient utilization of parallel processors for acceleration. Within convolutional layers, there are three types of operators: convolution used in forward propagation, deconvolution and dilated-convolution utilized in backward propagation. During the execution of these operators, zeros are typically added to tensors, leading to redundant calculations and unnecessary strain on hardware. To circumvent these inefficiencies, we propose the C-K-S algorithm, accompanied by efficient GPU implementations. C-K-S trims filters to exclude zero-padding. For deconvolution and dilated-convolution, C-K-S transforms sparse tensors into dense tensors, and standardizes the local computational rules to simplify the hardware control. The experimental results demonstrate that C-K-S offers good performance in terms of speed and convergence, surpassing the capabilities of PyTorch and cuDNN in certain scenarios.
[ { "created": "Wed, 28 Jun 2023 06:21:22 GMT", "version": "v1" }, { "created": "Wed, 12 Jul 2023 08:18:30 GMT", "version": "v2" }, { "created": "Sun, 5 Nov 2023 12:51:53 GMT", "version": "v3" } ]
2023-11-07
[ [ "Zhang", "Zhiyi", "" ], [ "Zhang", "Pengfei", "" ], [ "Xu", "Zhuopin", "" ], [ "Wang", "Qi", "" ] ]
Convolutional neural networks necessitate good algorithms to reduce complexity, and sufficient utilization of parallel processors for acceleration. Within convolutional layers, there are three types of operators: convolution used in forward propagation, deconvolution and dilated-convolution utilized in backward propagation. During the execution of these operators, zeros are typically added to tensors, leading to redundant calculations and unnecessary strain on hardware. To circumvent these inefficiencies, we propose the C-K-S algorithm, accompanied by efficient GPU implementations. C-K-S trims filters to exclude zero-padding. For deconvolution and dilated-convolution, C-K-S transforms sparse tensors into dense tensors, and standardizes the local computational rules to simplify the hardware control. The experimental results demonstrate that C-K-S offers good performance in terms of speed and convergence, surpassing the capabilities of PyTorch and cuDNN in certain scenarios.
0806.0172
Grenville Croll
David Chadwick
EuSpRIG TEAM work:Tools, Education, Audit, Management
7 Pages, 1 Figure
Proc. European Spreadsheet Risks Int. Grp. (EuSpRIG) 2003 1-6 ISBN 1 86166 199 1
null
null
cs.HC cs.CY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Research on spreadsheet errors began over fifteen years ago. During that time, there has been ample evidence demonstrating that spreadsheet errors are common and nontrivial. Quite simply, spreadsheet error rates are comparable to error rates in other human cognitive activities and are caused by fundamental limitations in human cognition, not mere sloppiness. Nor does ordinary "being careful" eliminate errors or reduce them to acceptable levels.
[ { "created": "Sun, 1 Jun 2008 21:01:15 GMT", "version": "v1" } ]
2008-06-03
[ [ "Chadwick", "David", "" ] ]
Research on spreadsheet errors began over fifteen years ago. During that time, there has been ample evidence demonstrating that spreadsheet errors are common and nontrivial. Quite simply, spreadsheet error rates are comparable to error rates in other human cognitive activities and are caused by fundamental limitations in human cognition, not mere sloppiness. Nor does ordinary "being careful" eliminate errors or reduce them to acceptable levels.
2404.14183
Yuxia Wang
Yuxia Wang, Jonibek Mansurov, Petar Ivanov, Jinyan Su, Artem Shelmanov, Akim Tsvigun, Osama Mohammed Afzal, Tarek Mahmoud, Giovanni Puccetti, Thomas Arnold, Chenxi Whitehouse, Alham Fikri Aji, Nizar Habash, Iryna Gurevych, Preslav Nakov
SemEval-2024 Task 8: Multidomain, Multimodel and Multilingual Machine-Generated Text Detection
23 pages, 12 tables
Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024)
null
null
cs.CL
http://creativecommons.org/licenses/by/4.0/
We present the results and the main findings of SemEval-2024 Task 8: Multigenerator, Multidomain, and Multilingual Machine-Generated Text Detection. The task featured three subtasks. Subtask A is a binary classification task determining whether a text is written by a human or generated by a machine. This subtask has two tracks: a monolingual track focused solely on English texts and a multilingual track. Subtask B is to detect the exact source of a text, discerning whether it is written by a human or generated by a specific LLM. Subtask C aims to identify the changing point within a text, at which the authorship transitions from human to machine. The task attracted a large number of participants: subtask A monolingual (126), subtask A multilingual (59), subtask B (70), and subtask C (30). In this paper, we present the task, analyze the results, and discuss the system submissions and the methods they used. For all subtasks, the best systems used LLMs.
[ { "created": "Mon, 22 Apr 2024 13:56:07 GMT", "version": "v1" } ]
2024-04-23
[ [ "Wang", "Yuxia", "" ], [ "Mansurov", "Jonibek", "" ], [ "Ivanov", "Petar", "" ], [ "Su", "Jinyan", "" ], [ "Shelmanov", "Artem", "" ], [ "Tsvigun", "Akim", "" ], [ "Afzal", "Osama Mohammed", "" ], [ "Mahmoud", "Tarek", "" ], [ "Puccetti", "Giovanni", "" ], [ "Arnold", "Thomas", "" ], [ "Whitehouse", "Chenxi", "" ], [ "Aji", "Alham Fikri", "" ], [ "Habash", "Nizar", "" ], [ "Gurevych", "Iryna", "" ], [ "Nakov", "Preslav", "" ] ]
We present the results and the main findings of SemEval-2024 Task 8: Multigenerator, Multidomain, and Multilingual Machine-Generated Text Detection. The task featured three subtasks. Subtask A is a binary classification task determining whether a text is written by a human or generated by a machine. This subtask has two tracks: a monolingual track focused solely on English texts and a multilingual track. Subtask B is to detect the exact source of a text, discerning whether it is written by a human or generated by a specific LLM. Subtask C aims to identify the changing point within a text, at which the authorship transitions from human to machine. The task attracted a large number of participants: subtask A monolingual (126), subtask A multilingual (59), subtask B (70), and subtask C (30). In this paper, we present the task, analyze the results, and discuss the system submissions and the methods they used. For all subtasks, the best systems used LLMs.
2403.06201
Huanqi Yang
Huanqi Yang, Sijie Ji, Rucheng Wu, Weitao Xu
Are You Being Tracked? Discover the Power of Zero-Shot Trajectory Tracing with LLMs!
null
null
null
null
cs.CL cs.AI cs.HC cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
There is a burgeoning discussion around the capabilities of Large Language Models (LLMs) in acting as fundamental components that can be seamlessly incorporated into Artificial Intelligence of Things (AIoT) to interpret complex trajectories. This study introduces LLMTrack, a model that illustrates how LLMs can be leveraged for Zero-Shot Trajectory Recognition by employing a novel single-prompt technique that combines role-play and think step-by-step methodologies with unprocessed Inertial Measurement Unit (IMU) data. We evaluate the model using real-world datasets designed to challenge it with distinct trajectories characterized by indoor and outdoor scenarios. In both test scenarios, LLMTrack not only meets but exceeds the performance benchmarks set by traditional machine learning approaches and even contemporary state-of-the-art deep learning models, all without the requirement of training on specialized datasets. The results of our research suggest that, with strategically designed prompts, LLMs can tap into their extensive knowledge base and are well-equipped to analyze raw sensor data with remarkable effectiveness.
[ { "created": "Sun, 10 Mar 2024 12:50:35 GMT", "version": "v1" } ]
2024-03-12
[ [ "Yang", "Huanqi", "" ], [ "Ji", "Sijie", "" ], [ "Wu", "Rucheng", "" ], [ "Xu", "Weitao", "" ] ]
There is a burgeoning discussion around the capabilities of Large Language Models (LLMs) in acting as fundamental components that can be seamlessly incorporated into Artificial Intelligence of Things (AIoT) to interpret complex trajectories. This study introduces LLMTrack, a model that illustrates how LLMs can be leveraged for Zero-Shot Trajectory Recognition by employing a novel single-prompt technique that combines role-play and think step-by-step methodologies with unprocessed Inertial Measurement Unit (IMU) data. We evaluate the model using real-world datasets designed to challenge it with distinct trajectories characterized by indoor and outdoor scenarios. In both test scenarios, LLMTrack not only meets but exceeds the performance benchmarks set by traditional machine learning approaches and even contemporary state-of-the-art deep learning models, all without the requirement of training on specialized datasets. The results of our research suggest that, with strategically designed prompts, LLMs can tap into their extensive knowledge base and are well-equipped to analyze raw sensor data with remarkable effectiveness.
2307.02627
Jacqueline Harding
Jacqueline Harding
Proxy Selection in Transitive Proxy Voting
null
Social Choice and Welfare 58, 69-99 (2022)
10.1007/s00355-021-01345-8
null
cs.GT cs.MA econ.GN q-fin.EC
http://creativecommons.org/licenses/by-nc-nd/4.0/
Transitive proxy voting (or "liquid democracy") is a novel form of collective decision making, often framed as an attractive hybrid of direct and representative democracy. Although the ideas behind liquid democracy have garnered widespread support, there have been relatively few attempts to model it formally. This paper makes three main contributions. First, it proposes a new social choice-theoretic model of liquid democracy, which is distinguished by taking a richer formal perspective on the process by which a voter chooses a proxy. Second, it examines the model from an axiomatic perspective, proving (a) a proxy vote analogue of May's Theorem and (b) an impossibility result concerning monotonicity properties in a proxy vote setting. Third, it explores the topic of manipulation in transitive proxy votes. Two forms of manipulation specific to the proxy vote setting are defined, and it is shown that manipulation occurs in strictly more cases in proxy votes than in classical votes.
[ { "created": "Wed, 5 Jul 2023 19:59:06 GMT", "version": "v1" } ]
2023-07-07
[ [ "Harding", "Jacqueline", "" ] ]
Transitive proxy voting (or "liquid democracy") is a novel form of collective decision making, often framed as an attractive hybrid of direct and representative democracy. Although the ideas behind liquid democracy have garnered widespread support, there have been relatively few attempts to model it formally. This paper makes three main contributions. First, it proposes a new social choice-theoretic model of liquid democracy, which is distinguished by taking a richer formal perspective on the process by which a voter chooses a proxy. Second, it examines the model from an axiomatic perspective, proving (a) a proxy vote analogue of May's Theorem and (b) an impossibility result concerning monotonicity properties in a proxy vote setting. Third, it explores the topic of manipulation in transitive proxy votes. Two forms of manipulation specific to the proxy vote setting are defined, and it is shown that manipulation occurs in strictly more cases in proxy votes than in classical votes.
2001.00088
Andrew Perrault
Andrew Perrault, Fei Fang, Arunesh Sinha, Milind Tambe
AI for Social Impact: Learning and Planning in the Data-to-Deployment Pipeline
AI Magazine, Winter 2020
null
10.1609/aimag.v41i4.5296
null
cs.CY cs.GT cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. In pursuit of this goal of AI for Social Impact, we as AI researchers must go beyond improvements in computational methodology; it is important to step out in the field to demonstrate social impact. To this end, we focus on the problems of public safety and security, wildlife conservation, and public health in low-resource communities, and present research advances in multiagent systems to address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. We present case studies from our deployments around the world as well as lessons learned that we hope are of use to researchers who are interested in AI for Social Impact. In pushing this research agenda, we believe AI can indeed play an important role in fighting social injustice and improving society.
[ { "created": "Mon, 16 Dec 2019 18:10:56 GMT", "version": "v1" }, { "created": "Sun, 12 Jun 2022 18:23:50 GMT", "version": "v2" } ]
2022-06-14
[ [ "Perrault", "Andrew", "" ], [ "Fang", "Fei", "" ], [ "Sinha", "Arunesh", "" ], [ "Tambe", "Milind", "" ] ]
With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. In pursuit of this goal of AI for Social Impact, we as AI researchers must go beyond improvements in computational methodology; it is important to step out in the field to demonstrate social impact. To this end, we focus on the problems of public safety and security, wildlife conservation, and public health in low-resource communities, and present research advances in multiagent systems to address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. We present case studies from our deployments around the world as well as lessons learned that we hope are of use to researchers who are interested in AI for Social Impact. In pushing this research agenda, we believe AI can indeed play an important role in fighting social injustice and improving society.
cs/0510080
Joseph Y. Halpern
Peter D. Grunwald and Joseph Y. Halpern
When Ignorance is Bliss
In Proceedings of the Twentieth Conference on Uncertainty in AI, 2004, pp. 226-234
null
null
null
cs.AI cs.LG
null
It is commonly-accepted wisdom that more information is better, and that information should never be ignored. Here we argue, using both a Bayesian and a non-Bayesian analysis, that in some situations you are better off ignoring information if your uncertainty is represented by a set of probability measures. These include situations in which the information is relevant for the prediction task at hand. In the non-Bayesian analysis, we show how ignoring information avoids dilation, the phenomenon that additional pieces of information sometimes lead to an increase in uncertainty. In the Bayesian analysis, we show that for small sample sizes and certain prediction tasks, the Bayesian posterior based on a noninformative prior yields worse predictions than simply ignoring the given information.
[ { "created": "Tue, 25 Oct 2005 22:14:33 GMT", "version": "v1" } ]
2007-05-23
[ [ "Grunwald", "Peter D.", "" ], [ "Halpern", "Joseph Y.", "" ] ]
It is commonly-accepted wisdom that more information is better, and that information should never be ignored. Here we argue, using both a Bayesian and a non-Bayesian analysis, that in some situations you are better off ignoring information if your uncertainty is represented by a set of probability measures. These include situations in which the information is relevant for the prediction task at hand. In the non-Bayesian analysis, we show how ignoring information avoids dilation, the phenomenon that additional pieces of information sometimes lead to an increase in uncertainty. In the Bayesian analysis, we show that for small sample sizes and certain prediction tasks, the Bayesian posterior based on a noninformative prior yields worse predictions than simply ignoring the given information.
1405.1894
Tillmann Miltzow
Michael Hoffmann, Vincent Kusters, Tillmann Miltzow
Halving Balls in Deterministic Linear Time
null
null
null
null
cs.CG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Let $\D$ be a set of $n$ pairwise disjoint unit balls in $\R^d$ and $P$ the set of their center points. A hyperplane $\Hy$ is an \emph{$m$-separator} for $\D$ if each closed halfspace bounded by $\Hy$ contains at least $m$ points from $P$. This generalizes the notion of halving hyperplanes, which correspond to $n/2$-separators. The analogous notion for point sets has been well studied. Separators have various applications, for instance, in divide-and-conquer schemes. In such a scheme any ball that is intersected by the separating hyperplane may still interact with both sides of the partition. Therefore it is desirable that the separating hyperplane intersects a small number of balls only. We present three deterministic algorithms to bisect or approximately bisect a given set of disjoint unit balls by a hyperplane: Firstly, we present a simple linear-time algorithm to construct an $\alpha n$-separator for balls in $\R^d$, for any $0<\alpha<1/2$, that intersects at most $cn^{(d-1)/d}$ balls, for some constant $c$ that depends on $d$ and $\alpha$. The number of intersected balls is best possible up to the constant $c$. Secondly, we present a near-linear time algorithm to construct an $(n/2-o(n))$-separator in $\R^d$ that intersects $o(n)$ balls. Finally, we give a linear-time algorithm to construct a halving line in $\R^2$ that intersects $O(n^{(5/6)+\epsilon})$ disks. Our results improve the runtime of a disk sliding algorithm by Bereg, Dumitrescu and Pach. In addition, our results improve and derandomize an algorithm to construct a space decomposition used by L{\"o}ffler and Mulzer to construct an onion (convex layer) decomposition for imprecise points (any point resides at an unknown location within a given disk).
[ { "created": "Thu, 8 May 2014 11:47:51 GMT", "version": "v1" } ]
2014-05-09
[ [ "Hoffmann", "Michael", "" ], [ "Kusters", "Vincent", "" ], [ "Miltzow", "Tillmann", "" ] ]
Let $\D$ be a set of $n$ pairwise disjoint unit balls in $\R^d$ and $P$ the set of their center points. A hyperplane $\Hy$ is an \emph{$m$-separator} for $\D$ if each closed halfspace bounded by $\Hy$ contains at least $m$ points from $P$. This generalizes the notion of halving hyperplanes, which correspond to $n/2$-separators. The analogous notion for point sets has been well studied. Separators have various applications, for instance, in divide-and-conquer schemes. In such a scheme any ball that is intersected by the separating hyperplane may still interact with both sides of the partition. Therefore it is desirable that the separating hyperplane intersects a small number of balls only. We present three deterministic algorithms to bisect or approximately bisect a given set of disjoint unit balls by a hyperplane: Firstly, we present a simple linear-time algorithm to construct an $\alpha n$-separator for balls in $\R^d$, for any $0<\alpha<1/2$, that intersects at most $cn^{(d-1)/d}$ balls, for some constant $c$ that depends on $d$ and $\alpha$. The number of intersected balls is best possible up to the constant $c$. Secondly, we present a near-linear time algorithm to construct an $(n/2-o(n))$-separator in $\R^d$ that intersects $o(n)$ balls. Finally, we give a linear-time algorithm to construct a halving line in $\R^2$ that intersects $O(n^{(5/6)+\epsilon})$ disks. Our results improve the runtime of a disk sliding algorithm by Bereg, Dumitrescu and Pach. In addition, our results improve and derandomize an algorithm to construct a space decomposition used by L{\"o}ffler and Mulzer to construct an onion (convex layer) decomposition for imprecise points (any point resides at an unknown location within a given disk).
1705.02116
Shuoyao Wang
Shuoyao Wang, Suzhi Bi, Ying Jun (Angela) Zhang, Jianwei Huang
Electrical Vehicle Charging Station Profit Maximization: Admission, Pricing, and Online Scheduling
This paper has been submitted to IEEE Transactions on Sustainable Energy for potential journal publication
null
null
null
cs.SY
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The rapid emergence of electric vehicles (EVs) demands an advanced infrastructure of publicly accessible charging stations that provide efficient charging services. In this paper, we propose a new charging station operation mechanism, the JoAP, which jointly optimizes the EV admission control, pricing, and charging scheduling to maximize the charging station's profit. More specifically, by introducing a tandem queueing network model, we analytically characterize the average charging station profit as a function of the admission control and pricing policies. Based on the analysis, we characterize the optimal JoAP algorithm. Through extensive simulations, we demonstrate that the proposed JoAP algorithm on average can achieve 330% and 531% higher profit than a widely adopted benchmark method under two representative waiting-time penalty rates.
[ { "created": "Fri, 5 May 2017 07:59:36 GMT", "version": "v1" }, { "created": "Thu, 7 Sep 2017 08:18:08 GMT", "version": "v2" } ]
2017-09-08
[ [ "Wang", "Shuoyao", "", "Angela" ], [ "Bi", "Suzhi", "", "Angela" ], [ "Jun", "Ying", "", "Angela" ], [ "Zhang", "", "" ], [ "Huang", "Jianwei", "" ] ]
The rapid emergence of electric vehicles (EVs) demands an advanced infrastructure of publicly accessible charging stations that provide efficient charging services. In this paper, we propose a new charging station operation mechanism, the JoAP, which jointly optimizes the EV admission control, pricing, and charging scheduling to maximize the charging station's profit. More specifically, by introducing a tandem queueing network model, we analytically characterize the average charging station profit as a function of the admission control and pricing policies. Based on the analysis, we characterize the optimal JoAP algorithm. Through extensive simulations, we demonstrate that the proposed JoAP algorithm on average can achieve 330% and 531% higher profit than a widely adopted benchmark method under two representative waiting-time penalty rates.
2012.08740
Yuhang Yao
Yuhang Yao, Carlee Joe-Wong
Interpretable Clustering on Dynamic Graphs with Recurrent Graph Neural Networks
AAAI 2021
AAAI 2021: 4608-4616
null
null
cs.LG cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study the problem of clustering nodes in a dynamic graph, where the connections between nodes and nodes' cluster memberships may change over time, e.g., due to community migration. We first propose a dynamic stochastic block model that captures these changes, and a simple decay-based clustering algorithm that clusters nodes based on weighted connections between them, where the weight decreases at a fixed rate over time. This decay rate can then be interpreted as signifying the importance of including historical connection information in the clustering. However, the optimal decay rate may differ for clusters with different rates of turnover. We characterize the optimal decay rate for each cluster and propose a clustering method that achieves almost exact recovery of the true clusters. We then demonstrate the efficacy of our clustering algorithm with optimized decay rates on simulated graph data. Recurrent neural networks (RNNs), a popular algorithm for sequence learning, use a similar decay-based method, and we use this insight to propose two new RNN-GCN (graph convolutional network) architectures for semi-supervised graph clustering. We finally demonstrate that the proposed architectures perform well on real data compared to state-of-the-art graph clustering algorithms.
[ { "created": "Wed, 16 Dec 2020 04:31:19 GMT", "version": "v1" }, { "created": "Tue, 22 Jun 2021 20:13:53 GMT", "version": "v2" } ]
2021-06-24
[ [ "Yao", "Yuhang", "" ], [ "Joe-Wong", "Carlee", "" ] ]
We study the problem of clustering nodes in a dynamic graph, where the connections between nodes and nodes' cluster memberships may change over time, e.g., due to community migration. We first propose a dynamic stochastic block model that captures these changes, and a simple decay-based clustering algorithm that clusters nodes based on weighted connections between them, where the weight decreases at a fixed rate over time. This decay rate can then be interpreted as signifying the importance of including historical connection information in the clustering. However, the optimal decay rate may differ for clusters with different rates of turnover. We characterize the optimal decay rate for each cluster and propose a clustering method that achieves almost exact recovery of the true clusters. We then demonstrate the efficacy of our clustering algorithm with optimized decay rates on simulated graph data. Recurrent neural networks (RNNs), a popular algorithm for sequence learning, use a similar decay-based method, and we use this insight to propose two new RNN-GCN (graph convolutional network) architectures for semi-supervised graph clustering. We finally demonstrate that the proposed architectures perform well on real data compared to state-of-the-art graph clustering algorithms.
1009.5346
Murugesan Kuttikrishnan
Murugesan Kuttikrishnan
A Novel Approach for Cardiac Disease Prediction and Classification Using Intelligent Agents
8 pages 2 figures and 7 tables
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 5, August 2010
null
null
cs.MA cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The goal is to develop a novel approach for cardiac disease prediction and diagnosis using intelligent agents. Initially the symptoms are preprocessed using filter and wrapper based agents. The filter removes the missing or irrelevant symptoms. Wrapper is used to extract the data in the data set according to the threshold limits. Dependency of each symptom is identified using dependency checker agent. The classification is based on the prior and posterior probability of the symptoms with the evidence value. Finally the symptoms are classified in to five classes namely absence, starting, mild, moderate and serious. Using the cooperative approach the cardiac problem is solved and verified.
[ { "created": "Mon, 27 Sep 2010 18:20:56 GMT", "version": "v1" } ]
2010-09-28
[ [ "Kuttikrishnan", "Murugesan", "" ] ]
The goal is to develop a novel approach for cardiac disease prediction and diagnosis using intelligent agents. Initially the symptoms are preprocessed using filter and wrapper based agents. The filter removes the missing or irrelevant symptoms. Wrapper is used to extract the data in the data set according to the threshold limits. Dependency of each symptom is identified using dependency checker agent. The classification is based on the prior and posterior probability of the symptoms with the evidence value. Finally the symptoms are classified in to five classes namely absence, starting, mild, moderate and serious. Using the cooperative approach the cardiac problem is solved and verified.
2302.05706
Wenxuan Wang
Wenxuan Wang, Jen-tse Huang, Weibin Wu, Jianping Zhang, Yizhan Huang, Shuqing Li, Pinjia He, Michael Lyu
MTTM: Metamorphic Testing for Textual Content Moderation Software
Accepted by ICSE 2023
null
null
null
cs.CL cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The exponential growth of social media platforms such as Twitter and Facebook has revolutionized textual communication and textual content publication in human society. However, they have been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography, which can lead to highly negative impacts (e.g., harmful effects on teen mental health). Researchers and practitioners have been enthusiastically developing and extensively deploying textual content moderation software to address this problem. However, we find that malicious users can evade moderation by changing only a few words in the toxic content. Moreover, modern content moderation software performance against malicious inputs remains underexplored. To this end, we propose MTTM, a Metamorphic Testing framework for Textual content Moderation software. Specifically, we conduct a pilot study on 2,000 text messages collected from real users and summarize eleven metamorphic relations across three perturbation levels: character, word, and sentence. MTTM employs these metamorphic relations on toxic textual contents to generate test cases, which are still toxic yet likely to evade moderation. In our evaluation, we employ MTTM to test three commercial textual content moderation software and two state-of-the-art moderation algorithms against three kinds of toxic content. The results show that MTTM achieves up to 83.9%, 51%, and 82.5% error finding rates (EFR) when testing commercial moderation software provided by Google, Baidu, and Huawei, respectively, and it obtains up to 91.2% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTTM to retrain the model we explored, which largely improves model robustness (0% to 5.9% EFR) while maintaining the accuracy on the original test set.
[ { "created": "Sat, 11 Feb 2023 14:44:39 GMT", "version": "v1" } ]
2023-02-14
[ [ "Wang", "Wenxuan", "" ], [ "Huang", "Jen-tse", "" ], [ "Wu", "Weibin", "" ], [ "Zhang", "Jianping", "" ], [ "Huang", "Yizhan", "" ], [ "Li", "Shuqing", "" ], [ "He", "Pinjia", "" ], [ "Lyu", "Michael", "" ] ]
The exponential growth of social media platforms such as Twitter and Facebook has revolutionized textual communication and textual content publication in human society. However, they have been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography, which can lead to highly negative impacts (e.g., harmful effects on teen mental health). Researchers and practitioners have been enthusiastically developing and extensively deploying textual content moderation software to address this problem. However, we find that malicious users can evade moderation by changing only a few words in the toxic content. Moreover, modern content moderation software performance against malicious inputs remains underexplored. To this end, we propose MTTM, a Metamorphic Testing framework for Textual content Moderation software. Specifically, we conduct a pilot study on 2,000 text messages collected from real users and summarize eleven metamorphic relations across three perturbation levels: character, word, and sentence. MTTM employs these metamorphic relations on toxic textual contents to generate test cases, which are still toxic yet likely to evade moderation. In our evaluation, we employ MTTM to test three commercial textual content moderation software and two state-of-the-art moderation algorithms against three kinds of toxic content. The results show that MTTM achieves up to 83.9%, 51%, and 82.5% error finding rates (EFR) when testing commercial moderation software provided by Google, Baidu, and Huawei, respectively, and it obtains up to 91.2% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTTM to retrain the model we explored, which largely improves model robustness (0% to 5.9% EFR) while maintaining the accuracy on the original test set.
2312.03173
Jaromir Savelka
Jacob Doughty, Zipiao Wan, Anishka Bompelli, Jubahed Qayum, Taozhi Wang, Juran Zhang, Yujia Zheng, Aidan Doyle, Pragnya Sridhar, Arav Agarwal, Christopher Bogart, Eric Keylor, Can Kultur, Jaromir Savelka, Majd Sakr
A Comparative Study of AI-Generated (GPT-4) and Human-crafted MCQs in Programming Education
null
null
10.1145/3636243.3636256
null
cs.CY cs.AI cs.CL
http://creativecommons.org/licenses/by/4.0/
There is a constant need for educators to develop and maintain effective up-to-date assessments. While there is a growing body of research in computing education on utilizing large language models (LLMs) in generation and engagement with coding exercises, the use of LLMs for generating programming MCQs has not been extensively explored. We analyzed the capability of GPT-4 to produce multiple-choice questions (MCQs) aligned with specific learning objectives (LOs) from Python programming classes in higher education. Specifically, we developed an LLM-powered (GPT-4) system for generation of MCQs from high-level course context and module-level LOs. We evaluated 651 LLM-generated and 449 human-crafted MCQs aligned to 246 LOs from 6 Python courses. We found that GPT-4 was capable of producing MCQs with clear language, a single correct choice, and high-quality distractors. We also observed that the generated MCQs appeared to be well-aligned with the LOs. Our findings can be leveraged by educators wishing to take advantage of the state-of-the-art generative models to support MCQ authoring efforts.
[ { "created": "Tue, 5 Dec 2023 22:29:43 GMT", "version": "v1" } ]
2023-12-07
[ [ "Doughty", "Jacob", "" ], [ "Wan", "Zipiao", "" ], [ "Bompelli", "Anishka", "" ], [ "Qayum", "Jubahed", "" ], [ "Wang", "Taozhi", "" ], [ "Zhang", "Juran", "" ], [ "Zheng", "Yujia", "" ], [ "Doyle", "Aidan", "" ], [ "Sridhar", "Pragnya", "" ], [ "Agarwal", "Arav", "" ], [ "Bogart", "Christopher", "" ], [ "Keylor", "Eric", "" ], [ "Kultur", "Can", "" ], [ "Savelka", "Jaromir", "" ], [ "Sakr", "Majd", "" ] ]
There is a constant need for educators to develop and maintain effective up-to-date assessments. While there is a growing body of research in computing education on utilizing large language models (LLMs) in generation and engagement with coding exercises, the use of LLMs for generating programming MCQs has not been extensively explored. We analyzed the capability of GPT-4 to produce multiple-choice questions (MCQs) aligned with specific learning objectives (LOs) from Python programming classes in higher education. Specifically, we developed an LLM-powered (GPT-4) system for generation of MCQs from high-level course context and module-level LOs. We evaluated 651 LLM-generated and 449 human-crafted MCQs aligned to 246 LOs from 6 Python courses. We found that GPT-4 was capable of producing MCQs with clear language, a single correct choice, and high-quality distractors. We also observed that the generated MCQs appeared to be well-aligned with the LOs. Our findings can be leveraged by educators wishing to take advantage of the state-of-the-art generative models to support MCQ authoring efforts.
2210.05896
Zhijie Wang
Shuangzhi Li, Zhijie Wang, Felix Juefei-Xu, Qing Guo, Xingyu Li and Lei Ma
Common Corruption Robustness of Point Cloud Detectors: Benchmark and Enhancement
16 pages, 6 figures
null
null
null
cs.CV cs.LG cs.RO
http://creativecommons.org/licenses/by/4.0/
Object detection through LiDAR-based point cloud has recently been important in autonomous driving. Although achieving high accuracy on public benchmarks, the state-of-the-art detectors may still go wrong and cause a heavy loss due to the widespread corruptions in the real world like rain, snow, sensor noise, etc. Nevertheless, there is a lack of a large-scale dataset covering diverse scenes and realistic corruption types with different severities to develop practical and robust point cloud detectors, which is challenging due to the heavy collection costs. To alleviate the challenge and start the first step for robust point cloud detection, we propose the physical-aware simulation methods to generate degraded point clouds under different real-world common corruptions. Then, for the first attempt, we construct a benchmark based on the physical-aware common corruptions for point cloud detectors, which contains a total of 1,122,150 examples covering 7,481 scenes, 25 common corruption types, and 6 severities. With such a novel benchmark, we conduct extensive empirical studies on 8 state-of-the-art detectors that contain 6 different detection frameworks. Thus we get several insight observations revealing the vulnerabilities of the detectors and indicating the enhancement directions. Moreover, we further study the effectiveness of existing robustness enhancement methods based on data augmentation and data denoising. The benchmark can potentially be a new platform for evaluating point cloud detectors, opening a door for developing novel robustness enhancement methods.
[ { "created": "Wed, 12 Oct 2022 03:23:35 GMT", "version": "v1" } ]
2022-10-13
[ [ "Li", "Shuangzhi", "" ], [ "Wang", "Zhijie", "" ], [ "Juefei-Xu", "Felix", "" ], [ "Guo", "Qing", "" ], [ "Li", "Xingyu", "" ], [ "Ma", "Lei", "" ] ]
Object detection through LiDAR-based point cloud has recently been important in autonomous driving. Although achieving high accuracy on public benchmarks, the state-of-the-art detectors may still go wrong and cause a heavy loss due to the widespread corruptions in the real world like rain, snow, sensor noise, etc. Nevertheless, there is a lack of a large-scale dataset covering diverse scenes and realistic corruption types with different severities to develop practical and robust point cloud detectors, which is challenging due to the heavy collection costs. To alleviate the challenge and start the first step for robust point cloud detection, we propose the physical-aware simulation methods to generate degraded point clouds under different real-world common corruptions. Then, for the first attempt, we construct a benchmark based on the physical-aware common corruptions for point cloud detectors, which contains a total of 1,122,150 examples covering 7,481 scenes, 25 common corruption types, and 6 severities. With such a novel benchmark, we conduct extensive empirical studies on 8 state-of-the-art detectors that contain 6 different detection frameworks. Thus we get several insight observations revealing the vulnerabilities of the detectors and indicating the enhancement directions. Moreover, we further study the effectiveness of existing robustness enhancement methods based on data augmentation and data denoising. The benchmark can potentially be a new platform for evaluating point cloud detectors, opening a door for developing novel robustness enhancement methods.
1012.4870
Ying Ding
Erjia Yan, Ying Ding (School of Library and Information Science, Indiana University, Bloomington, IN, United States)
Discovering author impact: A PageRank perspective
17 pages, 5 figures
null
null
null
cs.DL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This article provides an alternative perspective for measuring author impact by applying PageRank algorithm to a coauthorship network. A weighted PageRank algorithm considering citation and coauthorship network topology is proposed. We test this algorithm under different damping factors by evaluating author impact in the informetrics research community. In addition, we also compare this weighted PageRank with the h-index, citation, and program committee (PC) membership of the International Society for Scientometrics and Informetrics (ISSI) conferences. Findings show that this weighted PageRank algorithm provides reliable results in measuring author impact.
[ { "created": "Wed, 22 Dec 2010 03:05:20 GMT", "version": "v1" } ]
2010-12-23
[ [ "Yan", "Erjia", "", "School of Library and Information Science,\n Indiana University, Bloomington, IN, United States" ], [ "Ding", "Ying", "", "School of Library and Information Science,\n Indiana University, Bloomington, IN, United States" ] ]
This article provides an alternative perspective for measuring author impact by applying PageRank algorithm to a coauthorship network. A weighted PageRank algorithm considering citation and coauthorship network topology is proposed. We test this algorithm under different damping factors by evaluating author impact in the informetrics research community. In addition, we also compare this weighted PageRank with the h-index, citation, and program committee (PC) membership of the International Society for Scientometrics and Informetrics (ISSI) conferences. Findings show that this weighted PageRank algorithm provides reliable results in measuring author impact.
2206.07875
Risheng Liu
Risheng Liu, Xuan Liu, Shangzhi Zeng, Jin Zhang and Yixuan Zhang
Optimization-Derived Learning with Essential Convergence Analysis of Training and Hyper-training
Accepted by ICML 2022
null
null
null
cs.LG
http://creativecommons.org/licenses/by-sa/4.0/
Recently, Optimization-Derived Learning (ODL) has attracted attention from learning and vision areas, which designs learning models from the perspective of optimization. However, previous ODL approaches regard the training and hyper-training procedures as two separated stages, meaning that the hyper-training variables have to be fixed during the training process, and thus it is also impossible to simultaneously obtain the convergence of training and hyper-training variables. In this work, we design a Generalized Krasnoselskii-Mann (GKM) scheme based on fixed-point iterations as our fundamental ODL module, which unifies existing ODL methods as special cases. Under the GKM scheme, a Bilevel Meta Optimization (BMO) algorithmic framework is constructed to solve the optimal training and hyper-training variables together. We rigorously prove the essential joint convergence of the fixed-point iteration for training and the process of optimizing hyper-parameters for hyper-training, both on the approximation quality, and on the stationary analysis. Experiments demonstrate the efficiency of BMO with competitive performance on sparse coding and real-world applications such as image deconvolution and rain streak removal.
[ { "created": "Thu, 16 Jun 2022 01:50:25 GMT", "version": "v1" } ]
2022-06-17
[ [ "Liu", "Risheng", "" ], [ "Liu", "Xuan", "" ], [ "Zeng", "Shangzhi", "" ], [ "Zhang", "Jin", "" ], [ "Zhang", "Yixuan", "" ] ]
Recently, Optimization-Derived Learning (ODL) has attracted attention from learning and vision areas, which designs learning models from the perspective of optimization. However, previous ODL approaches regard the training and hyper-training procedures as two separated stages, meaning that the hyper-training variables have to be fixed during the training process, and thus it is also impossible to simultaneously obtain the convergence of training and hyper-training variables. In this work, we design a Generalized Krasnoselskii-Mann (GKM) scheme based on fixed-point iterations as our fundamental ODL module, which unifies existing ODL methods as special cases. Under the GKM scheme, a Bilevel Meta Optimization (BMO) algorithmic framework is constructed to solve the optimal training and hyper-training variables together. We rigorously prove the essential joint convergence of the fixed-point iteration for training and the process of optimizing hyper-parameters for hyper-training, both on the approximation quality, and on the stationary analysis. Experiments demonstrate the efficiency of BMO with competitive performance on sparse coding and real-world applications such as image deconvolution and rain streak removal.
1710.11381
Yannic Kilcher
Yannic Kilcher, Aurelien Lucchi, Thomas Hofmann
Semantic Interpolation in Implicit Models
null
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In implicit models, one often interpolates between sampled points in latent space. As we show in this paper, care needs to be taken to match-up the distributional assumptions on code vectors with the geometry of the interpolating paths. Otherwise, typical assumptions about the quality and semantics of in-between points may not be justified. Based on our analysis we propose to modify the prior code distribution to put significantly more probability mass closer to the origin. As a result, linear interpolation paths are not only shortest paths, but they are also guaranteed to pass through high-density regions, irrespective of the dimensionality of the latent space. Experiments on standard benchmark image datasets demonstrate clear visual improvements in the quality of the generated samples and exhibit more meaningful interpolation paths.
[ { "created": "Tue, 31 Oct 2017 09:11:17 GMT", "version": "v1" }, { "created": "Wed, 3 Jan 2018 08:56:11 GMT", "version": "v2" }, { "created": "Fri, 2 Feb 2018 09:56:08 GMT", "version": "v3" } ]
2018-02-05
[ [ "Kilcher", "Yannic", "" ], [ "Lucchi", "Aurelien", "" ], [ "Hofmann", "Thomas", "" ] ]
In implicit models, one often interpolates between sampled points in latent space. As we show in this paper, care needs to be taken to match-up the distributional assumptions on code vectors with the geometry of the interpolating paths. Otherwise, typical assumptions about the quality and semantics of in-between points may not be justified. Based on our analysis we propose to modify the prior code distribution to put significantly more probability mass closer to the origin. As a result, linear interpolation paths are not only shortest paths, but they are also guaranteed to pass through high-density regions, irrespective of the dimensionality of the latent space. Experiments on standard benchmark image datasets demonstrate clear visual improvements in the quality of the generated samples and exhibit more meaningful interpolation paths.
2205.06632
In\^es Terrucha
In\^es Terrucha, Elias Fern\'andez Domingos, Francisco C. Santos, Pieter Simoens and Tom Lenaerts
The art of compensation: how hybrid teams solve collective risk dilemmas
8 pages, 5 figures, accepted at workshop ALA 2022 (AAMAS 2022)
null
null
null
cs.CY
http://creativecommons.org/licenses/by/4.0/
It is widely known how the human ability to cooperate has influenced the thriving of our species. However, as we move towards a hybrid human-machine future, it is still unclear how the introduction of AI agents in our social interactions will affect this cooperative capacity. Within the context of the one-shot collective risk dilemma, where enough members of a group must cooperate in order to avoid a collective disaster, we study the evolutionary dynamics of cooperation in a hybrid population made of both adaptive and fixed-behavior agents. Specifically, we show how the first learn to adapt their behavior to compensate for the behavior of the latter. The less the (artificially) fixed agents cooperate, the more the adaptive population is motivated to cooperate, and vice-versa, especially when the risk is higher. By pinpointing how adaptive agents avoid their share of costly cooperation if the fixed-behavior agents implement a cooperative policy, our work hints towards an unbalanced hybrid world. On one hand, this means that introducing cooperative AI agents within our society might unburden human efforts. Nevertheless, it is important to note that costless artificial cooperation might not be realistic, and more than deploying AI systems that carry the cooperative effort, we must focus on mechanisms that nudge shared cooperation among all members in the hybrid system.
[ { "created": "Fri, 13 May 2022 13:23:42 GMT", "version": "v1" } ]
2022-05-16
[ [ "Terrucha", "Inês", "" ], [ "Domingos", "Elias Fernández", "" ], [ "Santos", "Francisco C.", "" ], [ "Simoens", "Pieter", "" ], [ "Lenaerts", "Tom", "" ] ]
It is widely known how the human ability to cooperate has influenced the thriving of our species. However, as we move towards a hybrid human-machine future, it is still unclear how the introduction of AI agents in our social interactions will affect this cooperative capacity. Within the context of the one-shot collective risk dilemma, where enough members of a group must cooperate in order to avoid a collective disaster, we study the evolutionary dynamics of cooperation in a hybrid population made of both adaptive and fixed-behavior agents. Specifically, we show how the first learn to adapt their behavior to compensate for the behavior of the latter. The less the (artificially) fixed agents cooperate, the more the adaptive population is motivated to cooperate, and vice-versa, especially when the risk is higher. By pinpointing how adaptive agents avoid their share of costly cooperation if the fixed-behavior agents implement a cooperative policy, our work hints towards an unbalanced hybrid world. On one hand, this means that introducing cooperative AI agents within our society might unburden human efforts. Nevertheless, it is important to note that costless artificial cooperation might not be realistic, and more than deploying AI systems that carry the cooperative effort, we must focus on mechanisms that nudge shared cooperation among all members in the hybrid system.
2112.04163
Jiayi Guo
Jiayi Guo, Chaoqun Du, Jiangshan Wang, Huijuan Huang, Pengfei Wan, Gao Huang
Assessing a Single Image in Reference-Guided Image Synthesis
Accepted by AAAI 2022
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Assessing the performance of Generative Adversarial Networks (GANs) has been an important topic due to its practical significance. Although several evaluation metrics have been proposed, they generally assess the quality of the whole generated image distribution. For Reference-guided Image Synthesis (RIS) tasks, i.e., rendering a source image in the style of another reference image, where assessing the quality of a single generated image is crucial, these metrics are not applicable. In this paper, we propose a general learning-based framework, Reference-guided Image Synthesis Assessment (RISA) to quantitatively evaluate the quality of a single generated image. Notably, the training of RISA does not require human annotations. In specific, the training data for RISA are acquired by the intermediate models from the training procedure in RIS, and weakly annotated by the number of models' iterations, based on the positive correlation between image quality and iterations. As this annotation is too coarse as a supervision signal, we introduce two techniques: 1) a pixel-wise interpolation scheme to refine the coarse labels, and 2) multiple binary classifiers to replace a na\"ive regressor. In addition, an unsupervised contrastive loss is introduced to effectively capture the style similarity between a generated image and its reference image. Empirical results on various datasets demonstrate that RISA is highly consistent with human preference and transfers well across models.
[ { "created": "Wed, 8 Dec 2021 08:22:14 GMT", "version": "v1" } ]
2021-12-09
[ [ "Guo", "Jiayi", "" ], [ "Du", "Chaoqun", "" ], [ "Wang", "Jiangshan", "" ], [ "Huang", "Huijuan", "" ], [ "Wan", "Pengfei", "" ], [ "Huang", "Gao", "" ] ]
Assessing the performance of Generative Adversarial Networks (GANs) has been an important topic due to its practical significance. Although several evaluation metrics have been proposed, they generally assess the quality of the whole generated image distribution. For Reference-guided Image Synthesis (RIS) tasks, i.e., rendering a source image in the style of another reference image, where assessing the quality of a single generated image is crucial, these metrics are not applicable. In this paper, we propose a general learning-based framework, Reference-guided Image Synthesis Assessment (RISA) to quantitatively evaluate the quality of a single generated image. Notably, the training of RISA does not require human annotations. In specific, the training data for RISA are acquired by the intermediate models from the training procedure in RIS, and weakly annotated by the number of models' iterations, based on the positive correlation between image quality and iterations. As this annotation is too coarse as a supervision signal, we introduce two techniques: 1) a pixel-wise interpolation scheme to refine the coarse labels, and 2) multiple binary classifiers to replace a na\"ive regressor. In addition, an unsupervised contrastive loss is introduced to effectively capture the style similarity between a generated image and its reference image. Empirical results on various datasets demonstrate that RISA is highly consistent with human preference and transfers well across models.
1910.10451
Manuel Steve Mbankeu Patchou
Manuel Patchou and Benjamin Sliwa and Christian Wietfeld
Unmanned Aerial Vehicles in Logistics: Efficiency Gains and Communication Performance of Hybrid Combinations of Ground and Aerial Vehicles
null
null
null
null
cs.NI eess.SP
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Unmanned Aerial Vehicles (UAVs) have drastically gained popularity in various Intelligent Transportation System (ITS) applications to improve the safety and efficiency of transportation systems. In this context, the combination of ground vehicles, such as delivery trucks, with drones to assist in the last mile pick-up and delivery of the parcels has been recently proposed. While aerial vehicles promise increased efficiency based on flexible routes and parallelized operation, highly reliable wireless communication is also required for the control and coordination of potentially many drones acting in a self-organized way. In this paper, we analyze the improvements procured by drone usage in parcel delivery compared to traditional delivery and propose a simulation framework to further quantify the efficiency gains of the parcel delivery logistics and to analyze the performance of different wireless communications options. To this end, we consider a heterogeneous vehicle routing problem with various constraints. We consider two approaches regarding the dispatching and recovery of drones and evaluate their benefits as opposed to parcel delivery with a classic truck only. Furthermore, we compare two networking technologies for enabling coordination of the self-organizing teams of drones with a realistically modeled environment: one approach relying on base station oriented Long Term Evolution (LTE) vs. a more decentralized Cellular Vehicle-to-Everything (C-V2X) solution. The results show time savings of nearly 40% can be achieved through drone usage and that the negative impact of urban shadowing on network communications in the base station oriented LTE approach can be compensated by leveraging decentralized C-V2X communications
[ { "created": "Wed, 23 Oct 2019 10:35:42 GMT", "version": "v1" }, { "created": "Thu, 7 Nov 2019 14:42:37 GMT", "version": "v2" } ]
2019-11-11
[ [ "Patchou", "Manuel", "" ], [ "Sliwa", "Benjamin", "" ], [ "Wietfeld", "Christian", "" ] ]
Unmanned Aerial Vehicles (UAVs) have drastically gained popularity in various Intelligent Transportation System (ITS) applications to improve the safety and efficiency of transportation systems. In this context, the combination of ground vehicles, such as delivery trucks, with drones to assist in the last mile pick-up and delivery of the parcels has been recently proposed. While aerial vehicles promise increased efficiency based on flexible routes and parallelized operation, highly reliable wireless communication is also required for the control and coordination of potentially many drones acting in a self-organized way. In this paper, we analyze the improvements procured by drone usage in parcel delivery compared to traditional delivery and propose a simulation framework to further quantify the efficiency gains of the parcel delivery logistics and to analyze the performance of different wireless communications options. To this end, we consider a heterogeneous vehicle routing problem with various constraints. We consider two approaches regarding the dispatching and recovery of drones and evaluate their benefits as opposed to parcel delivery with a classic truck only. Furthermore, we compare two networking technologies for enabling coordination of the self-organizing teams of drones with a realistically modeled environment: one approach relying on base station oriented Long Term Evolution (LTE) vs. a more decentralized Cellular Vehicle-to-Everything (C-V2X) solution. The results show time savings of nearly 40% can be achieved through drone usage and that the negative impact of urban shadowing on network communications in the base station oriented LTE approach can be compensated by leveraging decentralized C-V2X communications
0712.2638
Steve Oudot
Fr\'ed\'eric Chazal (INRIA Sophia Antipolis), Steve Oudot (INRIA Sophia Antipolis)
Towards Persistence-Based Reconstruction in Euclidean Spaces
null
null
null
null
cs.CG math.AT
null
Manifold reconstruction has been extensively studied for the last decade or so, especially in two and three dimensions. Recently, significant improvements were made in higher dimensions, leading to new methods to reconstruct large classes of compact subsets of Euclidean space $\R^d$. However, the complexities of these methods scale up exponentially with d, which makes them impractical in medium or high dimensions, even for handling low-dimensional submanifolds. In this paper, we introduce a novel approach that stands in-between classical reconstruction and topological estimation, and whose complexity scales up with the intrinsic dimension of the data. Specifically, when the data points are sufficiently densely sampled from a smooth $m$-submanifold of $\R^d$, our method retrieves the homology of the submanifold in time at most $c(m)n^5$, where $n$ is the size of the input and $c(m)$ is a constant depending solely on $m$. It can also provably well handle a wide range of compact subsets of $\R^d$, though with worse complexities. Along the way to proving the correctness of our algorithm, we obtain new results on \v{C}ech, Rips, and witness complex filtrations in Euclidean spaces.
[ { "created": "Mon, 17 Dec 2007 06:30:08 GMT", "version": "v1" }, { "created": "Tue, 18 Dec 2007 10:26:34 GMT", "version": "v2" } ]
2007-12-18
[ [ "Chazal", "Frédéric", "", "INRIA Sophia Antipolis" ], [ "Oudot", "Steve", "", "INRIA\n Sophia Antipolis" ] ]
Manifold reconstruction has been extensively studied for the last decade or so, especially in two and three dimensions. Recently, significant improvements were made in higher dimensions, leading to new methods to reconstruct large classes of compact subsets of Euclidean space $\R^d$. However, the complexities of these methods scale up exponentially with d, which makes them impractical in medium or high dimensions, even for handling low-dimensional submanifolds. In this paper, we introduce a novel approach that stands in-between classical reconstruction and topological estimation, and whose complexity scales up with the intrinsic dimension of the data. Specifically, when the data points are sufficiently densely sampled from a smooth $m$-submanifold of $\R^d$, our method retrieves the homology of the submanifold in time at most $c(m)n^5$, where $n$ is the size of the input and $c(m)$ is a constant depending solely on $m$. It can also provably well handle a wide range of compact subsets of $\R^d$, though with worse complexities. Along the way to proving the correctness of our algorithm, we obtain new results on \v{C}ech, Rips, and witness complex filtrations in Euclidean spaces.
2406.10415
Terrence Neumann
Terrence Neumann and Bryan Jones
PRISM: A Design Framework for Open-Source Foundation Model Safety
null
null
null
null
cs.CY cs.AI cs.SE
http://creativecommons.org/licenses/by/4.0/
The rapid advancement of open-source foundation models has brought transparency and accessibility to this groundbreaking technology. However, this openness has also enabled the development of highly-capable, unsafe models, as exemplified by recent instances such as WormGPT and FraudGPT, which are specifically designed to facilitate criminal activity. As the capabilities of open foundation models continue to grow, potentially outpacing those of closed-source models, the risk of misuse by bad actors poses an increasingly serious threat to society. This paper addresses the critical question of how open foundation model developers should approach model safety in light of these challenges. Our analysis reveals that open-source foundation model companies often provide less restrictive acceptable use policies (AUPs) compared to their closed-source counterparts, likely due to the inherent difficulties in enforcing such policies once the models are released. To tackle this issue, we introduce PRISM, a design framework for open-source foundation model safety that emphasizes Private, Robust, Independent Safety measures, at Minimal marginal cost of compute. The PRISM framework proposes the use of modular functions that moderate prompts and outputs independently of the core language model, offering a more adaptable and resilient approach to safety compared to the brittle reinforcement learning methods currently used for value alignment. By focusing on identifying AUP violations and engaging the developer community in establishing consensus around safety design decisions, PRISM aims to create a safer open-source ecosystem that maximizes the potential of these powerful technologies while minimizing the risks to individuals and society as a whole.
[ { "created": "Fri, 14 Jun 2024 21:26:15 GMT", "version": "v1" } ]
2024-06-18
[ [ "Neumann", "Terrence", "" ], [ "Jones", "Bryan", "" ] ]
The rapid advancement of open-source foundation models has brought transparency and accessibility to this groundbreaking technology. However, this openness has also enabled the development of highly-capable, unsafe models, as exemplified by recent instances such as WormGPT and FraudGPT, which are specifically designed to facilitate criminal activity. As the capabilities of open foundation models continue to grow, potentially outpacing those of closed-source models, the risk of misuse by bad actors poses an increasingly serious threat to society. This paper addresses the critical question of how open foundation model developers should approach model safety in light of these challenges. Our analysis reveals that open-source foundation model companies often provide less restrictive acceptable use policies (AUPs) compared to their closed-source counterparts, likely due to the inherent difficulties in enforcing such policies once the models are released. To tackle this issue, we introduce PRISM, a design framework for open-source foundation model safety that emphasizes Private, Robust, Independent Safety measures, at Minimal marginal cost of compute. The PRISM framework proposes the use of modular functions that moderate prompts and outputs independently of the core language model, offering a more adaptable and resilient approach to safety compared to the brittle reinforcement learning methods currently used for value alignment. By focusing on identifying AUP violations and engaging the developer community in establishing consensus around safety design decisions, PRISM aims to create a safer open-source ecosystem that maximizes the potential of these powerful technologies while minimizing the risks to individuals and society as a whole.
2205.01818
Ziyi Yang
Ziyi Yang, Yuwei Fang, Chenguang Zhu, Reid Pryzant, Dongdong Chen, Yu Shi, Yichong Xu, Yao Qian, Mei Gao, Yi-Ling Chen, Liyang Lu, Yujia Xie, Robert Gmyr, Noel Codella, Naoyuki Kanda, Bin Xiao, Lu Yuan, Takuya Yoshioka, Michael Zeng, Xuedong Huang
i-Code: An Integrative and Composable Multimodal Learning Framework
null
null
null
null
cs.LG cs.AI cs.CL cs.CV eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Human intelligence is multimodal; we integrate visual, linguistic, and acoustic signals to maintain a holistic worldview. Most current pretraining methods, however, are limited to one or two modalities. We present i-Code, a self-supervised pretraining framework where users may flexibly combine the modalities of vision, speech, and language into unified and general-purpose vector representations. In this framework, data from each modality are first given to pretrained single-modality encoders. The encoder outputs are then integrated with a multimodal fusion network, which uses novel attention mechanisms and other architectural innovations to effectively combine information from the different modalities. The entire system is pretrained end-to-end with new objectives including masked modality unit modeling and cross-modality contrastive learning. Unlike previous research using only video for pretraining, the i-Code framework can dynamically process single, dual, and triple-modality data during training and inference, flexibly projecting different combinations of modalities into a single representation space. Experimental results demonstrate how i-Code can outperform state-of-the-art techniques on five video understanding tasks and the GLUE NLP benchmark, improving by as much as 11% and demonstrating the power of integrative multimodal pretraining.
[ { "created": "Tue, 3 May 2022 23:38:50 GMT", "version": "v1" }, { "created": "Thu, 5 May 2022 06:35:23 GMT", "version": "v2" } ]
2022-05-06
[ [ "Yang", "Ziyi", "" ], [ "Fang", "Yuwei", "" ], [ "Zhu", "Chenguang", "" ], [ "Pryzant", "Reid", "" ], [ "Chen", "Dongdong", "" ], [ "Shi", "Yu", "" ], [ "Xu", "Yichong", "" ], [ "Qian", "Yao", "" ], [ "Gao", "Mei", "" ], [ "Chen", "Yi-Ling", "" ], [ "Lu", "Liyang", "" ], [ "Xie", "Yujia", "" ], [ "Gmyr", "Robert", "" ], [ "Codella", "Noel", "" ], [ "Kanda", "Naoyuki", "" ], [ "Xiao", "Bin", "" ], [ "Yuan", "Lu", "" ], [ "Yoshioka", "Takuya", "" ], [ "Zeng", "Michael", "" ], [ "Huang", "Xuedong", "" ] ]
Human intelligence is multimodal; we integrate visual, linguistic, and acoustic signals to maintain a holistic worldview. Most current pretraining methods, however, are limited to one or two modalities. We present i-Code, a self-supervised pretraining framework where users may flexibly combine the modalities of vision, speech, and language into unified and general-purpose vector representations. In this framework, data from each modality are first given to pretrained single-modality encoders. The encoder outputs are then integrated with a multimodal fusion network, which uses novel attention mechanisms and other architectural innovations to effectively combine information from the different modalities. The entire system is pretrained end-to-end with new objectives including masked modality unit modeling and cross-modality contrastive learning. Unlike previous research using only video for pretraining, the i-Code framework can dynamically process single, dual, and triple-modality data during training and inference, flexibly projecting different combinations of modalities into a single representation space. Experimental results demonstrate how i-Code can outperform state-of-the-art techniques on five video understanding tasks and the GLUE NLP benchmark, improving by as much as 11% and demonstrating the power of integrative multimodal pretraining.
1007.1593
Yakov Nekrich
Yakov Nekrich
A Fast Algorithm for Three-Dimensional Layers of Maxima Problem
null
null
null
null
cs.DS cs.CG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We show that the three-dimensional layers-of-maxima problem can be solved in $o(n\log n)$ time in the word RAM model. Our algorithm runs in $O(n(\log \log n)^3)$ deterministic time or $O(n(\log\log n)^2)$ expected time and uses O(n) space. We also describe an algorithm that uses optimal O(n) space and solves the three-dimensional layers-of-maxima problem in $O(n\log n)$ time in the pointer machine model.
[ { "created": "Fri, 9 Jul 2010 13:45:05 GMT", "version": "v1" }, { "created": "Tue, 3 May 2011 13:08:33 GMT", "version": "v2" } ]
2011-05-04
[ [ "Nekrich", "Yakov", "" ] ]
We show that the three-dimensional layers-of-maxima problem can be solved in $o(n\log n)$ time in the word RAM model. Our algorithm runs in $O(n(\log \log n)^3)$ deterministic time or $O(n(\log\log n)^2)$ expected time and uses O(n) space. We also describe an algorithm that uses optimal O(n) space and solves the three-dimensional layers-of-maxima problem in $O(n\log n)$ time in the pointer machine model.
2103.09191
Luigi Antonio Lavazza
Luigi Lavazza and Sandro Morasca
Understanding and Modeling AI-Intensive System Development
null
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Developers of AI-Intensive Systems--i.e., systems that involve both "traditional" software and Artificial Intelligence"are recognizing the need to organize development systematically and use engineered methods and tools. Since an AI-Intensive System (AIIS) relies heavily on software, it is expected that Software Engineering (SE) methods and tools can help. However, AIIS development differs from the development of "traditional" software systems in a few substantial aspects. Hence, traditional SE methods and tools are not suitable or sufficient by themselves and need to be adapted and extended. A quest for "SE for AI" methods and tools has started. We believe that, in this effort, we should learn from experience and avoid repeating some of the mistakes made in the quest for SE in past years. To this end, a fundamental instrument is a set of concepts and a notation to deal with AIIS and the problems that characterize their development processes. In this paper, we propose to describe AIIS via a notation that was proposed for SE and embeds a set of concepts that are suitable to represent AIIS as well. We demonstrate the usage of the notation by modeling some characteristics that are particularly relevant for AIIS.
[ { "created": "Tue, 16 Mar 2021 16:42:45 GMT", "version": "v1" } ]
2021-03-17
[ [ "Lavazza", "Luigi", "" ], [ "Morasca", "Sandro", "" ] ]
Developers of AI-Intensive Systems--i.e., systems that involve both "traditional" software and Artificial Intelligence"are recognizing the need to organize development systematically and use engineered methods and tools. Since an AI-Intensive System (AIIS) relies heavily on software, it is expected that Software Engineering (SE) methods and tools can help. However, AIIS development differs from the development of "traditional" software systems in a few substantial aspects. Hence, traditional SE methods and tools are not suitable or sufficient by themselves and need to be adapted and extended. A quest for "SE for AI" methods and tools has started. We believe that, in this effort, we should learn from experience and avoid repeating some of the mistakes made in the quest for SE in past years. To this end, a fundamental instrument is a set of concepts and a notation to deal with AIIS and the problems that characterize their development processes. In this paper, we propose to describe AIIS via a notation that was proposed for SE and embeds a set of concepts that are suitable to represent AIIS as well. We demonstrate the usage of the notation by modeling some characteristics that are particularly relevant for AIIS.
2408.01091
Jin Gao
Jin Gao, Lei Gan, Yuankai Li, Yixin Ye, Dequan Wang
Dissecting Dissonance: Benchmarking Large Multimodal Models Against Self-Contradictory Instructions
Accepted by the 18th European Conference on Computer Vision ECCV 2024
null
null
null
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Large multimodal models (LMMs) excel in adhering to human instructions. However, self-contradictory instructions may arise due to the increasing trend of multimodal interaction and context length, which is challenging for language beginners and vulnerable populations. We introduce the Self-Contradictory Instructions benchmark to evaluate the capability of LMMs in recognizing conflicting commands. It comprises 20,000 conflicts, evenly distributed between language and vision paradigms. It is constructed by a novel automatic dataset creation framework, which expedites the process and enables us to encompass a wide range of instruction forms. Our comprehensive evaluation reveals current LMMs consistently struggle to identify multimodal instruction discordance due to a lack of self-awareness. Hence, we propose the Cognitive Awakening Prompting to inject cognition from external, largely enhancing dissonance detection. The dataset and code are here: https://selfcontradiction.github.io/.
[ { "created": "Fri, 2 Aug 2024 08:11:11 GMT", "version": "v1" }, { "created": "Mon, 5 Aug 2024 06:56:44 GMT", "version": "v2" } ]
2024-08-06
[ [ "Gao", "Jin", "" ], [ "Gan", "Lei", "" ], [ "Li", "Yuankai", "" ], [ "Ye", "Yixin", "" ], [ "Wang", "Dequan", "" ] ]
Large multimodal models (LMMs) excel in adhering to human instructions. However, self-contradictory instructions may arise due to the increasing trend of multimodal interaction and context length, which is challenging for language beginners and vulnerable populations. We introduce the Self-Contradictory Instructions benchmark to evaluate the capability of LMMs in recognizing conflicting commands. It comprises 20,000 conflicts, evenly distributed between language and vision paradigms. It is constructed by a novel automatic dataset creation framework, which expedites the process and enables us to encompass a wide range of instruction forms. Our comprehensive evaluation reveals current LMMs consistently struggle to identify multimodal instruction discordance due to a lack of self-awareness. Hence, we propose the Cognitive Awakening Prompting to inject cognition from external, largely enhancing dissonance detection. The dataset and code are here: https://selfcontradiction.github.io/.
1509.01756
Xueru Li
Xueru Li, Emil Bj\"ornson, Erik G. Larsson, Shidong Zhou, Jing Wang
A Multi-cell MMSE Detector for Massive MIMO Systems and New Large System Analysis
6 pages, 3 figures, accepted by Globecom 2015
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, a new multi-cell MMSE detector is proposed for massive MIMO systems. Let $K$ and $B$ denote the number of users in each cell and the number of available pilot sequences in the network, respectively, with $B = \beta K$, where $\beta \ge 1 $ is called the pilot reuse factor. The novelty of the multi-cell MMSE detector is that it utilizes all $B$ channel directions that can be estimated locally at a base station, so that intra-cell interference, parts of the inter-cell interference and the noise can all be actively suppressed, while conventional detectors only use the $K$ intra-cell channels. Furthermore, in the large-system limit, a deterministic equivalent expression of the uplink SINR for the proposed multi-cell MMSE is derived. The expression is easy to compute and accounts for power control for the pilot and payload, imperfect channel estimation and arbitrary pilot allocation. Numerical results show that significant sum spectral efficiency gains can be obtained by the multi-cell MMSE over the conventional single-cell MMSE and the recent multi-cell ZF, and the gains become more significant as $\beta$ and/or $K$ increases. Furthermore, the deterministic equivalent is shown to be very accurate even for relatively small system dimensions.
[ { "created": "Sun, 6 Sep 2015 02:06:27 GMT", "version": "v1" } ]
2015-09-08
[ [ "Li", "Xueru", "" ], [ "Björnson", "Emil", "" ], [ "Larsson", "Erik G.", "" ], [ "Zhou", "Shidong", "" ], [ "Wang", "Jing", "" ] ]
In this paper, a new multi-cell MMSE detector is proposed for massive MIMO systems. Let $K$ and $B$ denote the number of users in each cell and the number of available pilot sequences in the network, respectively, with $B = \beta K$, where $\beta \ge 1 $ is called the pilot reuse factor. The novelty of the multi-cell MMSE detector is that it utilizes all $B$ channel directions that can be estimated locally at a base station, so that intra-cell interference, parts of the inter-cell interference and the noise can all be actively suppressed, while conventional detectors only use the $K$ intra-cell channels. Furthermore, in the large-system limit, a deterministic equivalent expression of the uplink SINR for the proposed multi-cell MMSE is derived. The expression is easy to compute and accounts for power control for the pilot and payload, imperfect channel estimation and arbitrary pilot allocation. Numerical results show that significant sum spectral efficiency gains can be obtained by the multi-cell MMSE over the conventional single-cell MMSE and the recent multi-cell ZF, and the gains become more significant as $\beta$ and/or $K$ increases. Furthermore, the deterministic equivalent is shown to be very accurate even for relatively small system dimensions.
2203.05683
Mayur Mallya
Mayur Mallya and Ghassan Hamarneh
Deep Multimodal Guidance for Medical Image Classification
null
null
null
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Medical imaging is a cornerstone of therapy and diagnosis in modern medicine. However, the choice of imaging modality for a particular theranostic task typically involves trade-offs between the feasibility of using a particular modality (e.g., short wait times, low cost, fast acquisition, reduced radiation/invasiveness) and the expected performance on a clinical task (e.g., diagnostic accuracy, efficacy of treatment planning and guidance). In this work, we aim to apply the knowledge learned from the less feasible but better-performing (superior) modality to guide the utilization of the more-feasible yet under-performing (inferior) modality and steer it towards improved performance. We focus on the application of deep learning for image-based diagnosis. We develop a light-weight guidance model that leverages the latent representation learned from the superior modality, when training a model that consumes only the inferior modality. We examine the advantages of our method in the context of two clinical applications: multi-task skin lesion classification from clinical and dermoscopic images and brain tumor classification from multi-sequence magnetic resonance imaging (MRI) and histopathology images. For both these scenarios we show a boost in diagnostic performance of the inferior modality without requiring the superior modality. Furthermore, in the case of brain tumor classification, our method outperforms the model trained on the superior modality while producing comparable results to the model that uses both modalities during inference.
[ { "created": "Thu, 10 Mar 2022 23:50:08 GMT", "version": "v1" }, { "created": "Thu, 21 Jul 2022 15:41:26 GMT", "version": "v2" } ]
2022-07-22
[ [ "Mallya", "Mayur", "" ], [ "Hamarneh", "Ghassan", "" ] ]
Medical imaging is a cornerstone of therapy and diagnosis in modern medicine. However, the choice of imaging modality for a particular theranostic task typically involves trade-offs between the feasibility of using a particular modality (e.g., short wait times, low cost, fast acquisition, reduced radiation/invasiveness) and the expected performance on a clinical task (e.g., diagnostic accuracy, efficacy of treatment planning and guidance). In this work, we aim to apply the knowledge learned from the less feasible but better-performing (superior) modality to guide the utilization of the more-feasible yet under-performing (inferior) modality and steer it towards improved performance. We focus on the application of deep learning for image-based diagnosis. We develop a light-weight guidance model that leverages the latent representation learned from the superior modality, when training a model that consumes only the inferior modality. We examine the advantages of our method in the context of two clinical applications: multi-task skin lesion classification from clinical and dermoscopic images and brain tumor classification from multi-sequence magnetic resonance imaging (MRI) and histopathology images. For both these scenarios we show a boost in diagnostic performance of the inferior modality without requiring the superior modality. Furthermore, in the case of brain tumor classification, our method outperforms the model trained on the superior modality while producing comparable results to the model that uses both modalities during inference.
2203.14531
Ye Zheng
Xiaoke Jiang, Donghai Li, Hao Chen, Ye Zheng, Rui Zhao and Liwei Wu
Uni6D: A Unified CNN Framework without Projection Breakdown for 6D Pose Estimation
CVPR2022
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
As RGB-D sensors become more affordable, using RGB-D images to obtain high-accuracy 6D pose estimation results becomes a better option. State-of-the-art approaches typically use different backbones to extract features for RGB and depth images. They use a 2D CNN for RGB images and a per-pixel point cloud network for depth data, as well as a fusion network for feature fusion. We find that the essential reason for using two independent backbones is the "projection breakdown" problem. In the depth image plane, the projected 3D structure of the physical world is preserved by the 1D depth value and its built-in 2D pixel coordinate (UV). Any spatial transformation that modifies UV, such as resize, flip, crop, or pooling operations in the CNN pipeline, breaks the binding between the pixel value and UV coordinate. As a consequence, the 3D structure is no longer preserved by a modified depth image or feature. To address this issue, we propose a simple yet effective method denoted as Uni6D that explicitly takes the extra UV data along with RGB-D images as input. Our method has a Unified CNN framework for 6D pose estimation with a single CNN backbone. In particular, the architecture of our method is based on Mask R-CNN with two extra heads, one named RT head for directly predicting 6D pose and the other named abc head for guiding the network to map the visible points to their coordinates in the 3D model as an auxiliary module. This end-to-end approach balances simplicity and accuracy, achieving comparable accuracy with state of the arts and 7.2x faster inference speed on the YCB-Video dataset.
[ { "created": "Mon, 28 Mar 2022 07:05:27 GMT", "version": "v1" }, { "created": "Tue, 5 Apr 2022 04:04:54 GMT", "version": "v2" } ]
2022-04-06
[ [ "Jiang", "Xiaoke", "" ], [ "Li", "Donghai", "" ], [ "Chen", "Hao", "" ], [ "Zheng", "Ye", "" ], [ "Zhao", "Rui", "" ], [ "Wu", "Liwei", "" ] ]
As RGB-D sensors become more affordable, using RGB-D images to obtain high-accuracy 6D pose estimation results becomes a better option. State-of-the-art approaches typically use different backbones to extract features for RGB and depth images. They use a 2D CNN for RGB images and a per-pixel point cloud network for depth data, as well as a fusion network for feature fusion. We find that the essential reason for using two independent backbones is the "projection breakdown" problem. In the depth image plane, the projected 3D structure of the physical world is preserved by the 1D depth value and its built-in 2D pixel coordinate (UV). Any spatial transformation that modifies UV, such as resize, flip, crop, or pooling operations in the CNN pipeline, breaks the binding between the pixel value and UV coordinate. As a consequence, the 3D structure is no longer preserved by a modified depth image or feature. To address this issue, we propose a simple yet effective method denoted as Uni6D that explicitly takes the extra UV data along with RGB-D images as input. Our method has a Unified CNN framework for 6D pose estimation with a single CNN backbone. In particular, the architecture of our method is based on Mask R-CNN with two extra heads, one named RT head for directly predicting 6D pose and the other named abc head for guiding the network to map the visible points to their coordinates in the 3D model as an auxiliary module. This end-to-end approach balances simplicity and accuracy, achieving comparable accuracy with state of the arts and 7.2x faster inference speed on the YCB-Video dataset.
1108.3525
Jason Corso
Yingjie Miao and Jason J. Corso
Hamiltonian Streamline Guided Feature Extraction with Applications to Face Detection
null
null
null
null
cs.CV math.DS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We propose a new feature extraction method based on two dynamical systems induced by intensity landscape: the negative gradient system and the Hamiltonian system. We build features based on the Hamiltonian streamlines. These features contain nice global topological information about the intensity landscape, and can be used for object detection. We show that for training images of same size, our feature space is much smaller than that generated by Haar-like features. The training time is extremely short, and detection speed and accuracy is similar to Haar-like feature based classifiers.
[ { "created": "Wed, 17 Aug 2011 17:06:41 GMT", "version": "v1" } ]
2011-08-18
[ [ "Miao", "Yingjie", "" ], [ "Corso", "Jason J.", "" ] ]
We propose a new feature extraction method based on two dynamical systems induced by intensity landscape: the negative gradient system and the Hamiltonian system. We build features based on the Hamiltonian streamlines. These features contain nice global topological information about the intensity landscape, and can be used for object detection. We show that for training images of same size, our feature space is much smaller than that generated by Haar-like features. The training time is extremely short, and detection speed and accuracy is similar to Haar-like feature based classifiers.
1910.06813
Anindya Sarkar
Anindya Sarkar, Anirudh Sunder Raj, Raghu Sesha Iyengar
ODE guided Neural Data Augmentation Techniques for Time Series Data and its Benefits on Robustness
8 pages, 5 figures, International Conference on Machine Learning and Applications
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Exploring adversarial attack vectors and studying their effects on machine learning algorithms has been of interest to researchers. Deep neural networks working with time series data have received lesser interest compared to their image counterparts in this context. In a recent finding, it has been revealed that current state-of-the-art deep learning time series classifiers are vulnerable to adversarial attacks. In this paper, we introduce two local gradient based and one spectral density based time series data augmentation techniques. We show that a model trained with data obtained using our techniques obtains state-of-the-art classification accuracy on various time series benchmarks. In addition, it improves the robustness of the model against some of the most common corruption techniques,such as Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM).
[ { "created": "Tue, 15 Oct 2019 14:37:18 GMT", "version": "v1" }, { "created": "Thu, 2 Jul 2020 10:31:48 GMT", "version": "v2" }, { "created": "Sun, 27 Sep 2020 17:53:53 GMT", "version": "v3" } ]
2020-09-29
[ [ "Sarkar", "Anindya", "" ], [ "Raj", "Anirudh Sunder", "" ], [ "Iyengar", "Raghu Sesha", "" ] ]
Exploring adversarial attack vectors and studying their effects on machine learning algorithms has been of interest to researchers. Deep neural networks working with time series data have received lesser interest compared to their image counterparts in this context. In a recent finding, it has been revealed that current state-of-the-art deep learning time series classifiers are vulnerable to adversarial attacks. In this paper, we introduce two local gradient based and one spectral density based time series data augmentation techniques. We show that a model trained with data obtained using our techniques obtains state-of-the-art classification accuracy on various time series benchmarks. In addition, it improves the robustness of the model against some of the most common corruption techniques,such as Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM).
2106.13731
Nestor Demeure
Less Wright and Nestor Demeure
Ranger21: a synergistic deep learning optimizer
for associated code, see https://github.com/lessw2020/Ranger21
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
As optimizers are critical to the performances of neural networks, every year a large number of papers innovating on the subject are published. However, while most of these publications provide incremental improvements to existing algorithms, they tend to be presented as new optimizers rather than composable algorithms. Thus, many worthwhile improvements are rarely seen out of their initial publication. Taking advantage of this untapped potential, we introduce Ranger21, a new optimizer which combines AdamW with eight components, carefully selected after reviewing and testing ideas from the literature. We found that the resulting optimizer provides significantly improved validation accuracy and training speed, smoother training curves, and is even able to train a ResNet50 on ImageNet2012 without Batch Normalization layers. A problem on which AdamW stays systematically stuck in a bad initial state.
[ { "created": "Fri, 25 Jun 2021 16:07:59 GMT", "version": "v1" }, { "created": "Sat, 7 Aug 2021 01:18:28 GMT", "version": "v2" } ]
2021-08-10
[ [ "Wright", "Less", "" ], [ "Demeure", "Nestor", "" ] ]
As optimizers are critical to the performances of neural networks, every year a large number of papers innovating on the subject are published. However, while most of these publications provide incremental improvements to existing algorithms, they tend to be presented as new optimizers rather than composable algorithms. Thus, many worthwhile improvements are rarely seen out of their initial publication. Taking advantage of this untapped potential, we introduce Ranger21, a new optimizer which combines AdamW with eight components, carefully selected after reviewing and testing ideas from the literature. We found that the resulting optimizer provides significantly improved validation accuracy and training speed, smoother training curves, and is even able to train a ResNet50 on ImageNet2012 without Batch Normalization layers. A problem on which AdamW stays systematically stuck in a bad initial state.
1903.00620
Jie Li
Jie Li, Yu Liu, Dong Gong, Qinfeng Shi, Xia Yuan, Chunxia Zhao, Ian Reid
RGBD Based Dimensional Decomposition Residual Network for 3D Semantic Scene Completion
CVPR2019
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
RGB images differentiate from depth images as they carry more details about the color and texture information, which can be utilized as a vital complementary to depth for boosting the performance of 3D semantic scene completion (SSC). SSC is composed of 3D shape completion (SC) and semantic scene labeling while most of the existing methods use depth as the sole input which causes the performance bottleneck. Moreover, the state-of-the-art methods employ 3D CNNs which have cumbersome networks and tremendous parameters. We introduce a light-weight Dimensional Decomposition Residual network (DDR) for 3D dense prediction tasks. The novel factorized convolution layer is effective for reducing the network parameters, and the proposed multi-scale fusion mechanism for depth and color image can improve the completion and segmentation accuracy simultaneously. Our method demonstrates excellent performance on two public datasets. Compared with the latest method SSCNet, we achieve 5.9% gains in SC-IoU and 5.7% gains in SSC-IOU, albeit with only 21% network parameters and 16.6% FLOPs employed compared with that of SSCNet.
[ { "created": "Sat, 2 Mar 2019 04:14:31 GMT", "version": "v1" }, { "created": "Wed, 1 May 2019 03:54:34 GMT", "version": "v2" } ]
2019-05-02
[ [ "Li", "Jie", "" ], [ "Liu", "Yu", "" ], [ "Gong", "Dong", "" ], [ "Shi", "Qinfeng", "" ], [ "Yuan", "Xia", "" ], [ "Zhao", "Chunxia", "" ], [ "Reid", "Ian", "" ] ]
RGB images differentiate from depth images as they carry more details about the color and texture information, which can be utilized as a vital complementary to depth for boosting the performance of 3D semantic scene completion (SSC). SSC is composed of 3D shape completion (SC) and semantic scene labeling while most of the existing methods use depth as the sole input which causes the performance bottleneck. Moreover, the state-of-the-art methods employ 3D CNNs which have cumbersome networks and tremendous parameters. We introduce a light-weight Dimensional Decomposition Residual network (DDR) for 3D dense prediction tasks. The novel factorized convolution layer is effective for reducing the network parameters, and the proposed multi-scale fusion mechanism for depth and color image can improve the completion and segmentation accuracy simultaneously. Our method demonstrates excellent performance on two public datasets. Compared with the latest method SSCNet, we achieve 5.9% gains in SC-IoU and 5.7% gains in SSC-IOU, albeit with only 21% network parameters and 16.6% FLOPs employed compared with that of SSCNet.
2204.04960
Adil Erzin I
Adil Erzin, Roman Plotnikov, Ilya Ladygin
Constrained Shortest Path and Hierarchical Structures
null
null
null
null
cs.DS cs.DM
http://creativecommons.org/licenses/by-nc-nd/4.0/
The Constraint Shortest Path (CSP) problem is as follows. An $n$-vertex graph is given, each edge/arc assigned two weights. Let us call them "cost" and "length" for definiteness. Finding a min-cost upper-bounded length path between a given pair of vertices is required. The problem is NP-hard even when the lengths of all edges are the same. Therefore, various approximation algorithms have been proposed in the literature for it. The constraint on path length can be accounted for by considering one edge weight equals to a linear combination of cost and length. By varying the multiplier value in a linear combination, a feasible solution delivers a minimum to the function with new weights. At the same time, Dijkstra's algorithm or its modifications are used to construct the shortest path with the current weights of the edges. However, with insufficiently large graphs, this approach may turn out to be time-consuming. In this article, we propose to look for a solution, not in the original graph but specially constructed hierarchical structures (HS). We show that the shortest path in the HS is constructed with $O(m)$-time complexity, where $m$ is the number of edges/arcs of the graph, and the approximate solution in the case of integer costs and lengths of the edges is found with $O(m\log n)$-time complexity. The a priori estimate of the algorithm's accuracy turned out to depend on the parameters of the problem and can be significant. Therefore, to evaluate the algorithm's effectiveness, we conducted a numerical experiment on the graphs of roads of megalopolis and randomly constructed unit-disk graphs (UDGs). The numerical experiment results show that in the HS, a solution close to optimal one is built 10--100 times faster than in the methods which use Dijkstra's algorithm to build a min-weight path in the original graph.
[ { "created": "Mon, 11 Apr 2022 09:13:43 GMT", "version": "v1" } ]
2022-04-12
[ [ "Erzin", "Adil", "" ], [ "Plotnikov", "Roman", "" ], [ "Ladygin", "Ilya", "" ] ]
The Constraint Shortest Path (CSP) problem is as follows. An $n$-vertex graph is given, each edge/arc assigned two weights. Let us call them "cost" and "length" for definiteness. Finding a min-cost upper-bounded length path between a given pair of vertices is required. The problem is NP-hard even when the lengths of all edges are the same. Therefore, various approximation algorithms have been proposed in the literature for it. The constraint on path length can be accounted for by considering one edge weight equals to a linear combination of cost and length. By varying the multiplier value in a linear combination, a feasible solution delivers a minimum to the function with new weights. At the same time, Dijkstra's algorithm or its modifications are used to construct the shortest path with the current weights of the edges. However, with insufficiently large graphs, this approach may turn out to be time-consuming. In this article, we propose to look for a solution, not in the original graph but specially constructed hierarchical structures (HS). We show that the shortest path in the HS is constructed with $O(m)$-time complexity, where $m$ is the number of edges/arcs of the graph, and the approximate solution in the case of integer costs and lengths of the edges is found with $O(m\log n)$-time complexity. The a priori estimate of the algorithm's accuracy turned out to depend on the parameters of the problem and can be significant. Therefore, to evaluate the algorithm's effectiveness, we conducted a numerical experiment on the graphs of roads of megalopolis and randomly constructed unit-disk graphs (UDGs). The numerical experiment results show that in the HS, a solution close to optimal one is built 10--100 times faster than in the methods which use Dijkstra's algorithm to build a min-weight path in the original graph.
1908.02723
Ian Fox
Ian Fox and Jenna Wiens
Advocacy Learning: Learning through Competition and Class-Conditional Representations
Accepted IJCAI 2019
null
null
null
cs.LG cs.CV stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce advocacy learning, a novel supervised training scheme for attention-based classification problems. Advocacy learning relies on a framework consisting of two connected networks: 1) $N$ Advocates (one for each class), each of which outputs an argument in the form of an attention map over the input, and 2) a Judge, which predicts the class label based on these arguments. Each Advocate produces a class-conditional representation with the goal of convincing the Judge that the input example belongs to their class, even when the input belongs to a different class. Applied to several different classification tasks, we show that advocacy learning can lead to small improvements in classification accuracy over an identical supervised baseline. Though a series of follow-up experiments, we analyze when and how such class-conditional representations improve discriminative performance. Though somewhat counter-intuitive, a framework in which subnetworks are trained to competitively provide evidence in support of their class shows promise, in many cases performing on par with standard learning approaches. This provides a foundation for further exploration into competition and class-conditional representations in supervised learning.
[ { "created": "Wed, 7 Aug 2019 16:55:44 GMT", "version": "v1" } ]
2019-08-08
[ [ "Fox", "Ian", "" ], [ "Wiens", "Jenna", "" ] ]
We introduce advocacy learning, a novel supervised training scheme for attention-based classification problems. Advocacy learning relies on a framework consisting of two connected networks: 1) $N$ Advocates (one for each class), each of which outputs an argument in the form of an attention map over the input, and 2) a Judge, which predicts the class label based on these arguments. Each Advocate produces a class-conditional representation with the goal of convincing the Judge that the input example belongs to their class, even when the input belongs to a different class. Applied to several different classification tasks, we show that advocacy learning can lead to small improvements in classification accuracy over an identical supervised baseline. Though a series of follow-up experiments, we analyze when and how such class-conditional representations improve discriminative performance. Though somewhat counter-intuitive, a framework in which subnetworks are trained to competitively provide evidence in support of their class shows promise, in many cases performing on par with standard learning approaches. This provides a foundation for further exploration into competition and class-conditional representations in supervised learning.
2003.02576
Antoine Amarilli
Antoine Amarilli, Pierre Bourhis, Stefan Mengel, Matthias Niewerth
Constant-Delay Enumeration for Nondeterministic Document Spanners
29 pages. Extended version of arXiv:1807.09320. Integrates all corrections following reviewer feedback. Outside of some minor formatting differences and tweaks, this paper is the same as the paper to appear in the ACM TODS journal
null
10.1145/3436487
null
cs.DB cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the information extraction framework known as document spanners, and study the problem of efficiently computing the results of the extraction from an input document, where the extraction task is described as a sequential variable-set automaton (VA). We pose this problem in the setting of enumeration algorithms, where we can first run a preprocessing phase and must then produce the results with a small delay between any two consecutive results. Our goal is to have an algorithm which is tractable in combined complexity, i.e., in the sizes of the input document and the VA; while ensuring the best possible data complexity bounds in the input document size, i.e., constant delay in the document size. Several recent works at PODS'18 proposed such algorithms but with linear delay in the document size or with an exponential dependency in size of the (generally nondeterministic) input VA. In particular, Florenzano et al. suggest that our desired runtime guarantees cannot be met for general sequential VAs. We refute this and show that, given a nondeterministic sequential VA and an input document, we can enumerate the mappings of the VA on the document with the following bounds: the preprocessing is linear in the document size and polynomial in the size of the VA, and the delay is independent of the document and polynomial in the size of the VA. The resulting algorithm thus achieves tractability in combined complexity and the best possible data complexity bounds. Moreover, it is rather easy to describe, in particular for the restricted case of so-called extended VAs. Finally, we evaluate our algorithm empirically using a prototype implementation.
[ { "created": "Thu, 5 Mar 2020 12:49:56 GMT", "version": "v1" }, { "created": "Fri, 25 Sep 2020 07:48:10 GMT", "version": "v2" }, { "created": "Mon, 7 Dec 2020 13:51:51 GMT", "version": "v3" } ]
2023-09-06
[ [ "Amarilli", "Antoine", "" ], [ "Bourhis", "Pierre", "" ], [ "Mengel", "Stefan", "" ], [ "Niewerth", "Matthias", "" ] ]
We consider the information extraction framework known as document spanners, and study the problem of efficiently computing the results of the extraction from an input document, where the extraction task is described as a sequential variable-set automaton (VA). We pose this problem in the setting of enumeration algorithms, where we can first run a preprocessing phase and must then produce the results with a small delay between any two consecutive results. Our goal is to have an algorithm which is tractable in combined complexity, i.e., in the sizes of the input document and the VA; while ensuring the best possible data complexity bounds in the input document size, i.e., constant delay in the document size. Several recent works at PODS'18 proposed such algorithms but with linear delay in the document size or with an exponential dependency in size of the (generally nondeterministic) input VA. In particular, Florenzano et al. suggest that our desired runtime guarantees cannot be met for general sequential VAs. We refute this and show that, given a nondeterministic sequential VA and an input document, we can enumerate the mappings of the VA on the document with the following bounds: the preprocessing is linear in the document size and polynomial in the size of the VA, and the delay is independent of the document and polynomial in the size of the VA. The resulting algorithm thus achieves tractability in combined complexity and the best possible data complexity bounds. Moreover, it is rather easy to describe, in particular for the restricted case of so-called extended VAs. Finally, we evaluate our algorithm empirically using a prototype implementation.
2204.01855
Shima Khoshraftar
Shima Khoshraftar, Aijun An
A Survey on Graph Representation Learning Methods
null
null
null
null
cs.LG cs.SI
http://creativecommons.org/publicdomain/zero/1.0/
Graphs representation learning has been a very active research area in recent years. The goal of graph representation learning is to generate graph representation vectors that capture the structure and features of large graphs accurately. This is especially important because the quality of the graph representation vectors will affect the performance of these vectors in downstream tasks such as node classification, link prediction and anomaly detection. Many techniques are proposed for generating effective graph representation vectors. Two of the most prevalent categories of graph representation learning are graph embedding methods without using graph neural nets (GNN), which we denote as non-GNN based graph embedding methods, and graph neural nets (GNN) based methods. Non-GNN graph embedding methods are based on techniques such as random walks, temporal point processes and neural network learning methods. GNN-based methods, on the other hand, are the application of deep learning on graph data. In this survey, we provide an overview of these two categories and cover the current state-of-the-art methods for both static and dynamic graphs. Finally, we explore some open and ongoing research directions for future work.
[ { "created": "Mon, 4 Apr 2022 21:18:48 GMT", "version": "v1" }, { "created": "Wed, 15 Jun 2022 17:26:31 GMT", "version": "v2" } ]
2022-06-16
[ [ "Khoshraftar", "Shima", "" ], [ "An", "Aijun", "" ] ]
Graphs representation learning has been a very active research area in recent years. The goal of graph representation learning is to generate graph representation vectors that capture the structure and features of large graphs accurately. This is especially important because the quality of the graph representation vectors will affect the performance of these vectors in downstream tasks such as node classification, link prediction and anomaly detection. Many techniques are proposed for generating effective graph representation vectors. Two of the most prevalent categories of graph representation learning are graph embedding methods without using graph neural nets (GNN), which we denote as non-GNN based graph embedding methods, and graph neural nets (GNN) based methods. Non-GNN graph embedding methods are based on techniques such as random walks, temporal point processes and neural network learning methods. GNN-based methods, on the other hand, are the application of deep learning on graph data. In this survey, we provide an overview of these two categories and cover the current state-of-the-art methods for both static and dynamic graphs. Finally, we explore some open and ongoing research directions for future work.
2310.10917
Boqun Zhao
Boqun Zhao, Chongjun Ouyang, Yuanwei Liu, Xingqi Zhang, H. Vincent Poor
Modeling and Analysis of Near-Field ISAC
Accepted by IEEE Journal of Selected Topics in Signal Processing
null
10.1109/JSTSP.2024.3386054
null
cs.IT eess.SP math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
As the technical trends for the next-generation wireless network significantly extend the near-field region, a performance reevaluation of integrated sensing and communications (ISAC) with an appropriate channel model to account for the effects introduced by the near field becomes essential. In this paper, a near-field ISAC framework is proposed for both downlink and uplink scenarios based on an accurate channel model. A uniform planar array is equipped at a base station, where the impacts of the effective aperture and polarization of antennas are considered. For the downlink case, three distinct designs are studied: a communications-centric (C-C) design, a sensing-centric (S-C) design, and a Pareto optimal design. Regarding the uplink case, the C-C design, the S-C design and a time-sharing strategy are considered. Within each design, sensing rates (SRs) and communication rates (CRs) are derived. To gain further insights, high signal-to-noise ratio slopes and rate scaling laws concerning the number of antennas are examined. The attainable near-field SR-CR regions of ISAC and the baseline frequency-division S&C are also characterized. Numerical results reveal that, as the number of antennas in the array grows, the SRs and CRs under our accurate model converge to finite values, while those under conventional far- and near-field models exhibit unbounded growth, highlighting the importance of precisely modeling the channels for near-field ISAC.
[ { "created": "Tue, 17 Oct 2023 01:25:23 GMT", "version": "v1" }, { "created": "Wed, 18 Oct 2023 20:05:53 GMT", "version": "v2" }, { "created": "Thu, 30 Nov 2023 15:17:27 GMT", "version": "v3" }, { "created": "Fri, 12 Apr 2024 23:26:36 GMT", "version": "v4" } ]
2024-04-16
[ [ "Zhao", "Boqun", "" ], [ "Ouyang", "Chongjun", "" ], [ "Liu", "Yuanwei", "" ], [ "Zhang", "Xingqi", "" ], [ "Poor", "H. Vincent", "" ] ]
As the technical trends for the next-generation wireless network significantly extend the near-field region, a performance reevaluation of integrated sensing and communications (ISAC) with an appropriate channel model to account for the effects introduced by the near field becomes essential. In this paper, a near-field ISAC framework is proposed for both downlink and uplink scenarios based on an accurate channel model. A uniform planar array is equipped at a base station, where the impacts of the effective aperture and polarization of antennas are considered. For the downlink case, three distinct designs are studied: a communications-centric (C-C) design, a sensing-centric (S-C) design, and a Pareto optimal design. Regarding the uplink case, the C-C design, the S-C design and a time-sharing strategy are considered. Within each design, sensing rates (SRs) and communication rates (CRs) are derived. To gain further insights, high signal-to-noise ratio slopes and rate scaling laws concerning the number of antennas are examined. The attainable near-field SR-CR regions of ISAC and the baseline frequency-division S&C are also characterized. Numerical results reveal that, as the number of antennas in the array grows, the SRs and CRs under our accurate model converge to finite values, while those under conventional far- and near-field models exhibit unbounded growth, highlighting the importance of precisely modeling the channels for near-field ISAC.
2102.13185
Zhuangdi Zhu
Zhuangdi Zhu, Kaixiang Lin, Bo Dai, Jiayu Zhou
Off-Policy Imitation Learning from Observations
null
null
null
null
cs.LG cs.AI
http://creativecommons.org/publicdomain/zero/1.0/
Learning from Observations (LfO) is a practical reinforcement learning scenario from which many applications can benefit through the reuse of incomplete resources. Compared to conventional imitation learning (IL), LfO is more challenging because of the lack of expert action guidance. In both conventional IL and LfO, distribution matching is at the heart of their foundation. Traditional distribution matching approaches are sample-costly which depend on on-policy transitions for policy learning. Towards sample-efficiency, some off-policy solutions have been proposed, which, however, either lack comprehensive theoretical justifications or depend on the guidance of expert actions. In this work, we propose a sample-efficient LfO approach that enables off-policy optimization in a principled manner. To further accelerate the learning procedure, we regulate the policy update with an inverse action model, which assists distribution matching from the perspective of mode-covering. Extensive empirical results on challenging locomotion tasks indicate that our approach is comparable with state-of-the-art in terms of both sample-efficiency and asymptotic performance.
[ { "created": "Thu, 25 Feb 2021 21:33:47 GMT", "version": "v1" } ]
2021-03-01
[ [ "Zhu", "Zhuangdi", "" ], [ "Lin", "Kaixiang", "" ], [ "Dai", "Bo", "" ], [ "Zhou", "Jiayu", "" ] ]
Learning from Observations (LfO) is a practical reinforcement learning scenario from which many applications can benefit through the reuse of incomplete resources. Compared to conventional imitation learning (IL), LfO is more challenging because of the lack of expert action guidance. In both conventional IL and LfO, distribution matching is at the heart of their foundation. Traditional distribution matching approaches are sample-costly which depend on on-policy transitions for policy learning. Towards sample-efficiency, some off-policy solutions have been proposed, which, however, either lack comprehensive theoretical justifications or depend on the guidance of expert actions. In this work, we propose a sample-efficient LfO approach that enables off-policy optimization in a principled manner. To further accelerate the learning procedure, we regulate the policy update with an inverse action model, which assists distribution matching from the perspective of mode-covering. Extensive empirical results on challenging locomotion tasks indicate that our approach is comparable with state-of-the-art in terms of both sample-efficiency and asymptotic performance.
1812.02937
Idoia Ruiz
Idoia Ruiz, Bogdan Raducanu, Rakesh Mehta, Jaume Amores
Optimizing speed/accuracy trade-off for person re-identification via knowledge distillation
Published on the journal "Engineering Applications of Artificial Intelligence"
Engineering Applications of Artificial Intelligence, Volume 87, January 2020, 103309
10.1016/j.engappai.2019.103309
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Finding a person across a camera network plays an important role in video surveillance. For a real-world person re-identification application, in order to guarantee an optimal time response, it is crucial to find the balance between accuracy and speed. We analyse this trade-off, comparing a classical method, that comprises hand-crafted feature description and metric learning, in particular, LOMO and XQDA, to deep learning based techniques, using image classification networks, ResNet and MobileNets. Additionally, we propose and analyse network distillation as a learning strategy to reduce the computational cost of the deep learning approach at test time. We evaluate both methods on the Market-1501 and DukeMTMC-reID large-scale datasets, showing that distillation helps reducing the computational cost at inference time while even increasing the accuracy performance.
[ { "created": "Fri, 7 Dec 2018 08:11:06 GMT", "version": "v1" }, { "created": "Thu, 5 Dec 2019 16:40:11 GMT", "version": "v2" } ]
2019-12-06
[ [ "Ruiz", "Idoia", "" ], [ "Raducanu", "Bogdan", "" ], [ "Mehta", "Rakesh", "" ], [ "Amores", "Jaume", "" ] ]
Finding a person across a camera network plays an important role in video surveillance. For a real-world person re-identification application, in order to guarantee an optimal time response, it is crucial to find the balance between accuracy and speed. We analyse this trade-off, comparing a classical method, that comprises hand-crafted feature description and metric learning, in particular, LOMO and XQDA, to deep learning based techniques, using image classification networks, ResNet and MobileNets. Additionally, we propose and analyse network distillation as a learning strategy to reduce the computational cost of the deep learning approach at test time. We evaluate both methods on the Market-1501 and DukeMTMC-reID large-scale datasets, showing that distillation helps reducing the computational cost at inference time while even increasing the accuracy performance.
2110.06195
Brian Cho
Brian Y. Cho, Tucker Hermans, Alan Kuntz
Planning Sensing Sequences for Subsurface 3D Tumor Mapping
7 pages, 9 figures, to be published in the proceedings of the 2021 International Symposium on Medical Robotics (ISMR)
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Surgical automation has the potential to enable increased precision and reduce the per-patient workload of overburdened human surgeons. An effective automation system must be able to sense and map subsurface anatomy, such as tumors, efficiently and accurately. In this work, we present a method that plans a sequence of sensing actions to map the 3D geometry of subsurface tumors. We leverage a sequential Bayesian Hilbert map to create a 3D probabilistic occupancy model that represents the likelihood that any given point in the anatomy is occupied by a tumor, conditioned on sensor readings. We iteratively update the map, utilizing Bayesian optimization to determine sensing poses that explore unsensed regions of anatomy and exploit the knowledge gained by previous sensing actions. We demonstrate our method's efficiency and accuracy in three anatomical scenarios including a liver tumor scenario generated from a real patient's CT scan. The results show that our proposed method significantly outperforms comparison methods in terms of efficiency while detecting subsurface tumors with high accuracy.
[ { "created": "Tue, 12 Oct 2021 17:48:41 GMT", "version": "v1" } ]
2021-10-13
[ [ "Cho", "Brian Y.", "" ], [ "Hermans", "Tucker", "" ], [ "Kuntz", "Alan", "" ] ]
Surgical automation has the potential to enable increased precision and reduce the per-patient workload of overburdened human surgeons. An effective automation system must be able to sense and map subsurface anatomy, such as tumors, efficiently and accurately. In this work, we present a method that plans a sequence of sensing actions to map the 3D geometry of subsurface tumors. We leverage a sequential Bayesian Hilbert map to create a 3D probabilistic occupancy model that represents the likelihood that any given point in the anatomy is occupied by a tumor, conditioned on sensor readings. We iteratively update the map, utilizing Bayesian optimization to determine sensing poses that explore unsensed regions of anatomy and exploit the knowledge gained by previous sensing actions. We demonstrate our method's efficiency and accuracy in three anatomical scenarios including a liver tumor scenario generated from a real patient's CT scan. The results show that our proposed method significantly outperforms comparison methods in terms of efficiency while detecting subsurface tumors with high accuracy.
2011.06964
Micha\"el Fanuel
Micha\"el Fanuel, Joachim Schreurs, Johan A.K. Suykens
Determinantal Point Processes Implicitly Regularize Semi-parametric Regression Problems
26 pages. Extended results. Typos corrected
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Semi-parametric regression models are used in several applications which require comprehensibility without sacrificing accuracy. Typical examples are spline interpolation in geophysics, or non-linear time series problems, where the system includes a linear and non-linear component. We discuss here the use of a finite Determinantal Point Process (DPP) for approximating semi-parametric models. Recently, Barthelm\'e, Tremblay, Usevich, and Amblard introduced a novel representation of some finite DPPs. These authors formulated extended L-ensembles that can conveniently represent partial-projection DPPs and suggest their use for optimal interpolation. With the help of this formalism, we derive a key identity illustrating the implicit regularization effect of determinantal sampling for semi-parametric regression and interpolation. Also, a novel projected Nystr\"om approximation is defined and used to derive a bound on the expected risk for the corresponding approximation of semi-parametric regression. This work naturally extends similar results obtained for kernel ridge regression.
[ { "created": "Fri, 13 Nov 2020 15:22:16 GMT", "version": "v1" }, { "created": "Tue, 9 Mar 2021 13:47:11 GMT", "version": "v2" } ]
2021-03-10
[ [ "Fanuel", "Michaël", "" ], [ "Schreurs", "Joachim", "" ], [ "Suykens", "Johan A. K.", "" ] ]
Semi-parametric regression models are used in several applications which require comprehensibility without sacrificing accuracy. Typical examples are spline interpolation in geophysics, or non-linear time series problems, where the system includes a linear and non-linear component. We discuss here the use of a finite Determinantal Point Process (DPP) for approximating semi-parametric models. Recently, Barthelm\'e, Tremblay, Usevich, and Amblard introduced a novel representation of some finite DPPs. These authors formulated extended L-ensembles that can conveniently represent partial-projection DPPs and suggest their use for optimal interpolation. With the help of this formalism, we derive a key identity illustrating the implicit regularization effect of determinantal sampling for semi-parametric regression and interpolation. Also, a novel projected Nystr\"om approximation is defined and used to derive a bound on the expected risk for the corresponding approximation of semi-parametric regression. This work naturally extends similar results obtained for kernel ridge regression.
2308.06998
Hao Shen
Hao Shen, Zhong-Qiu Zhao, Yulun Zhang, Zhao Zhang
Mutual Information-driven Triple Interaction Network for Efficient Image Dehazing
Accepted in ACM MM 2023
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Multi-stage architectures have exhibited efficacy in image dehazing, which usually decomposes a challenging task into multiple more tractable sub-tasks and progressively estimates latent hazy-free images. Despite the remarkable progress, existing methods still suffer from the following shortcomings: (1) limited exploration of frequency domain information; (2) insufficient information interaction; (3) severe feature redundancy. To remedy these issues, we propose a novel Mutual Information-driven Triple interaction Network (MITNet) based on spatial-frequency dual domain information and two-stage architecture. To be specific, the first stage, named amplitude-guided haze removal, aims to recover the amplitude spectrum of the hazy images for haze removal. And the second stage, named phase-guided structure refined, devotes to learning the transformation and refinement of the phase spectrum. To facilitate the information exchange between two stages, an Adaptive Triple Interaction Module (ATIM) is developed to simultaneously aggregate cross-domain, cross-scale, and cross-stage features, where the fused features are further used to generate content-adaptive dynamic filters so that applying them to enhance global context representation. In addition, we impose the mutual information minimization constraint on paired scale encoder and decoder features from both stages. Such an operation can effectively reduce information redundancy and enhance cross-stage feature complementarity. Extensive experiments on multiple public datasets exhibit that our MITNet performs superior performance with lower model complexity.The code and models are available at https://github.com/it-hao/MITNet.
[ { "created": "Mon, 14 Aug 2023 08:23:58 GMT", "version": "v1" } ]
2023-08-15
[ [ "Shen", "Hao", "" ], [ "Zhao", "Zhong-Qiu", "" ], [ "Zhang", "Yulun", "" ], [ "Zhang", "Zhao", "" ] ]
Multi-stage architectures have exhibited efficacy in image dehazing, which usually decomposes a challenging task into multiple more tractable sub-tasks and progressively estimates latent hazy-free images. Despite the remarkable progress, existing methods still suffer from the following shortcomings: (1) limited exploration of frequency domain information; (2) insufficient information interaction; (3) severe feature redundancy. To remedy these issues, we propose a novel Mutual Information-driven Triple interaction Network (MITNet) based on spatial-frequency dual domain information and two-stage architecture. To be specific, the first stage, named amplitude-guided haze removal, aims to recover the amplitude spectrum of the hazy images for haze removal. And the second stage, named phase-guided structure refined, devotes to learning the transformation and refinement of the phase spectrum. To facilitate the information exchange between two stages, an Adaptive Triple Interaction Module (ATIM) is developed to simultaneously aggregate cross-domain, cross-scale, and cross-stage features, where the fused features are further used to generate content-adaptive dynamic filters so that applying them to enhance global context representation. In addition, we impose the mutual information minimization constraint on paired scale encoder and decoder features from both stages. Such an operation can effectively reduce information redundancy and enhance cross-stage feature complementarity. Extensive experiments on multiple public datasets exhibit that our MITNet performs superior performance with lower model complexity.The code and models are available at https://github.com/it-hao/MITNet.
2404.08509
Haoran Qiu
Haoran Qiu, Weichao Mao, Archit Patke, Shengkun Cui, Saurabh Jha, Chen Wang, Hubertus Franke, Zbigniew T. Kalbarczyk, Tamer Ba\c{s}ar, Ravishankar K. Iyer
Efficient Interactive LLM Serving with Proxy Model-based Sequence Length Prediction
Accepted at AIOps'24
null
null
null
cs.DC cs.CL cs.LG
http://creativecommons.org/licenses/by/4.0/
Large language models (LLMs) have been driving a new wave of interactive AI applications across numerous domains. However, efficiently serving LLM inference requests is challenging due to their unpredictable execution times originating from the autoregressive nature of generative models. Existing LLM serving systems exploit first-come-first-serve (FCFS) scheduling, suffering from head-of-line blocking issues. To address the non-deterministic nature of LLMs and enable efficient interactive LLM serving, we present a speculative shortest-job-first (SSJF) scheduler that uses a light proxy model to predict LLM output sequence lengths. Our open-source SSJF implementation does not require changes to memory management or batching strategies. Evaluations on real-world datasets and production workload traces show that SSJF reduces average job completion times by 30.5-39.6% and increases throughput by 2.2-3.6x compared to FCFS schedulers, across no batching, dynamic batching, and continuous batching settings.
[ { "created": "Fri, 12 Apr 2024 14:46:15 GMT", "version": "v1" } ]
2024-04-15
[ [ "Qiu", "Haoran", "" ], [ "Mao", "Weichao", "" ], [ "Patke", "Archit", "" ], [ "Cui", "Shengkun", "" ], [ "Jha", "Saurabh", "" ], [ "Wang", "Chen", "" ], [ "Franke", "Hubertus", "" ], [ "Kalbarczyk", "Zbigniew T.", "" ], [ "Başar", "Tamer", "" ], [ "Iyer", "Ravishankar K.", "" ] ]
Large language models (LLMs) have been driving a new wave of interactive AI applications across numerous domains. However, efficiently serving LLM inference requests is challenging due to their unpredictable execution times originating from the autoregressive nature of generative models. Existing LLM serving systems exploit first-come-first-serve (FCFS) scheduling, suffering from head-of-line blocking issues. To address the non-deterministic nature of LLMs and enable efficient interactive LLM serving, we present a speculative shortest-job-first (SSJF) scheduler that uses a light proxy model to predict LLM output sequence lengths. Our open-source SSJF implementation does not require changes to memory management or batching strategies. Evaluations on real-world datasets and production workload traces show that SSJF reduces average job completion times by 30.5-39.6% and increases throughput by 2.2-3.6x compared to FCFS schedulers, across no batching, dynamic batching, and continuous batching settings.
1304.2714
Henry E. Kyburg Jr.
Henry E. Kyburg Jr
Higher Order Probabilities
Appears in Proceedings of the Third Conference on Uncertainty in Artificial Intelligence (UAI1987)
null
null
UAI-P-1987-PG-30-38
cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A number of writers have supposed that for the full specification of belief, higher order probabilities are required. Some have even supposed that there may be an unending sequence of higher order probabilities of probabilities of probabilities.... In the present paper we show that higher order probabilities can always be replaced by the marginal distributions of joint probability distributions. We consider both the case in which higher order probabilities are of the same sort as lower order probabilities and that in which higher order probabilities are distinct in character, as when lower order probabilities are construed as frequencies and higher order probabilities are construed as subjective degrees of belief. In neither case do higher order probabilities appear to offer any advantages, either conceptually or computationally.
[ { "created": "Wed, 27 Mar 2013 19:46:32 GMT", "version": "v1" } ]
2013-04-11
[ [ "Kyburg", "Henry E.", "Jr" ] ]
A number of writers have supposed that for the full specification of belief, higher order probabilities are required. Some have even supposed that there may be an unending sequence of higher order probabilities of probabilities of probabilities.... In the present paper we show that higher order probabilities can always be replaced by the marginal distributions of joint probability distributions. We consider both the case in which higher order probabilities are of the same sort as lower order probabilities and that in which higher order probabilities are distinct in character, as when lower order probabilities are construed as frequencies and higher order probabilities are construed as subjective degrees of belief. In neither case do higher order probabilities appear to offer any advantages, either conceptually or computationally.
2108.09241
Jie Huang
Jie Huang, Kevin Chen-Chuan Chang, Jinjun Xiong, Wen-mei Hwu
Open Relation Modeling: Learning to Define Relations between Entities
Accepted to Findings of ACL 2022
null
null
null
cs.CL cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Relations between entities can be represented by different instances, e.g., a sentence containing both entities or a fact in a Knowledge Graph (KG). However, these instances may not well capture the general relations between entities, may be difficult to understand by humans, even may not be found due to the incompleteness of the knowledge source. In this paper, we introduce the Open Relation Modeling problem - given two entities, generate a coherent sentence describing the relation between them. To solve this problem, we propose to teach machines to generate definition-like relation descriptions by letting them learn from defining entities. Specifically, we fine-tune Pre-trained Language Models (PLMs) to produce definitions conditioned on extracted entity pairs. To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling, we incorporate reasoning paths in KGs and include a reasoning path selection mechanism. Experimental results show that our model can generate concise but informative relation descriptions that capture the representative characteristics of entities.
[ { "created": "Fri, 20 Aug 2021 16:03:23 GMT", "version": "v1" }, { "created": "Thu, 3 Mar 2022 04:36:32 GMT", "version": "v2" } ]
2022-03-04
[ [ "Huang", "Jie", "" ], [ "Chang", "Kevin Chen-Chuan", "" ], [ "Xiong", "Jinjun", "" ], [ "Hwu", "Wen-mei", "" ] ]
Relations between entities can be represented by different instances, e.g., a sentence containing both entities or a fact in a Knowledge Graph (KG). However, these instances may not well capture the general relations between entities, may be difficult to understand by humans, even may not be found due to the incompleteness of the knowledge source. In this paper, we introduce the Open Relation Modeling problem - given two entities, generate a coherent sentence describing the relation between them. To solve this problem, we propose to teach machines to generate definition-like relation descriptions by letting them learn from defining entities. Specifically, we fine-tune Pre-trained Language Models (PLMs) to produce definitions conditioned on extracted entity pairs. To help PLMs reason between entities and provide additional relational knowledge to PLMs for open relation modeling, we incorporate reasoning paths in KGs and include a reasoning path selection mechanism. Experimental results show that our model can generate concise but informative relation descriptions that capture the representative characteristics of entities.
2106.07505
Vanessa Hahn
Vanessa Hahn, Dana Ruiter, Thomas Kleinbauer, Dietrich Klakow
Modeling Profanity and Hate Speech in Social Media with Semantic Subspaces
9 pages, 4 figures, accepted as a long paper at Workshop on Online Abuse and Harms 2021
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Hate speech and profanity detection suffer from data sparsity, especially for languages other than English, due to the subjective nature of the tasks and the resulting annotation incompatibility of existing corpora. In this study, we identify profane subspaces in word and sentence representations and explore their generalization capability on a variety of similar and distant target tasks in a zero-shot setting. This is done monolingually (German) and cross-lingually to closely-related (English), distantly-related (French) and non-related (Arabic) tasks. We observe that, on both similar and distant target tasks and across all languages, the subspace-based representations transfer more effectively than standard BERT representations in the zero-shot setting, with improvements between F1 +10.9 and F1 +42.9 over the baselines across all tested monolingual and cross-lingual scenarios.
[ { "created": "Mon, 14 Jun 2021 15:34:37 GMT", "version": "v1" }, { "created": "Fri, 18 Jun 2021 10:04:11 GMT", "version": "v2" } ]
2021-06-21
[ [ "Hahn", "Vanessa", "" ], [ "Ruiter", "Dana", "" ], [ "Kleinbauer", "Thomas", "" ], [ "Klakow", "Dietrich", "" ] ]
Hate speech and profanity detection suffer from data sparsity, especially for languages other than English, due to the subjective nature of the tasks and the resulting annotation incompatibility of existing corpora. In this study, we identify profane subspaces in word and sentence representations and explore their generalization capability on a variety of similar and distant target tasks in a zero-shot setting. This is done monolingually (German) and cross-lingually to closely-related (English), distantly-related (French) and non-related (Arabic) tasks. We observe that, on both similar and distant target tasks and across all languages, the subspace-based representations transfer more effectively than standard BERT representations in the zero-shot setting, with improvements between F1 +10.9 and F1 +42.9 over the baselines across all tested monolingual and cross-lingual scenarios.
1211.7113
Tinku Rasheed
Djamal-Eddine Meddour, Tinku Rasheed and Yvon Gourhant
On the Role of Infrastructure sharing for Mobile Network Operators in Emerging Markets
null
The International Journal of Computer and Telecommunications Networking, Volume 55, Issue 7, 2011, Pages 1576-1591
10.1016/j.comnet.2011.01.023
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The traditional model of single ownership of all the physical network elements and network layers by mobile network operators is beginning to be challenged. This has been attributed to the rapid and complex technology migration compounded with rigorous regulatory requirements and ever increasing capital expenditures. These trends, combined together with the increasing competition, rapid commoditization of telecommunication equipments and rising separation of network and service provisioning are pushing the operators to adopt multiple strategies, with network infrastructure sharing in the core and radio access networks emerging as a more radical mechanism to substantially and sustainably improve network costs. Through infrastructure sharing, developing countries and other emerging economies can harness the technological, market and regulatory developments that have fostered affordable access to mobile and broadband services. Similarly, the network operators entering or consolidating in the emerging markets can aim for substantial savings on capital and operating expenses. The present paper aims to investigate the current technological solutions and regulatory and the technical-economical dimensions in connection with the sharing of mobile telecommunication networks in emerging countries. We analyze the estimated savings on capital and operating expenses, while assessing the technical constraints, applicability and benefits of the network sharing solutions in an emerging market context.
[ { "created": "Thu, 29 Nov 2012 22:51:56 GMT", "version": "v1" } ]
2012-12-03
[ [ "Meddour", "Djamal-Eddine", "" ], [ "Rasheed", "Tinku", "" ], [ "Gourhant", "Yvon", "" ] ]
The traditional model of single ownership of all the physical network elements and network layers by mobile network operators is beginning to be challenged. This has been attributed to the rapid and complex technology migration compounded with rigorous regulatory requirements and ever increasing capital expenditures. These trends, combined together with the increasing competition, rapid commoditization of telecommunication equipments and rising separation of network and service provisioning are pushing the operators to adopt multiple strategies, with network infrastructure sharing in the core and radio access networks emerging as a more radical mechanism to substantially and sustainably improve network costs. Through infrastructure sharing, developing countries and other emerging economies can harness the technological, market and regulatory developments that have fostered affordable access to mobile and broadband services. Similarly, the network operators entering or consolidating in the emerging markets can aim for substantial savings on capital and operating expenses. The present paper aims to investigate the current technological solutions and regulatory and the technical-economical dimensions in connection with the sharing of mobile telecommunication networks in emerging countries. We analyze the estimated savings on capital and operating expenses, while assessing the technical constraints, applicability and benefits of the network sharing solutions in an emerging market context.
1309.2350
Shahin Shahrampour
Shahin Shahrampour and Ali Jadbabaie
Exponentially Fast Parameter Estimation in Networks Using Distributed Dual Averaging
6 pages, To appear in Conference on Decision and Control 2013
null
null
null
cs.LG cs.SI math.OC stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we present an optimization-based view of distributed parameter estimation and observational social learning in networks. Agents receive a sequence of random, independent and identically distributed (i.i.d.) signals, each of which individually may not be informative about the underlying true state, but the signals together are globally informative enough to make the true state identifiable. Using an optimization-based characterization of Bayesian learning as proximal stochastic gradient descent (with Kullback-Leibler divergence from a prior as a proximal function), we show how to efficiently use a distributed, online variant of Nesterov's dual averaging method to solve the estimation with purely local information. When the true state is globally identifiable, and the network is connected, we prove that agents eventually learn the true parameter using a randomized gossip scheme. We demonstrate that with high probability the convergence is exponentially fast with a rate dependent on the KL divergence of observations under the true state from observations under the second likeliest state. Furthermore, our work also highlights the possibility of learning under continuous adaptation of network which is a consequence of employing constant, unit stepsize for the algorithm.
[ { "created": "Tue, 10 Sep 2013 00:36:44 GMT", "version": "v1" } ]
2013-09-11
[ [ "Shahrampour", "Shahin", "" ], [ "Jadbabaie", "Ali", "" ] ]
In this paper we present an optimization-based view of distributed parameter estimation and observational social learning in networks. Agents receive a sequence of random, independent and identically distributed (i.i.d.) signals, each of which individually may not be informative about the underlying true state, but the signals together are globally informative enough to make the true state identifiable. Using an optimization-based characterization of Bayesian learning as proximal stochastic gradient descent (with Kullback-Leibler divergence from a prior as a proximal function), we show how to efficiently use a distributed, online variant of Nesterov's dual averaging method to solve the estimation with purely local information. When the true state is globally identifiable, and the network is connected, we prove that agents eventually learn the true parameter using a randomized gossip scheme. We demonstrate that with high probability the convergence is exponentially fast with a rate dependent on the KL divergence of observations under the true state from observations under the second likeliest state. Furthermore, our work also highlights the possibility of learning under continuous adaptation of network which is a consequence of employing constant, unit stepsize for the algorithm.
2309.16347
Eleftherios Triantafyllidis Mr.
Eleftherios Triantafyllidis, Filippos Christianos and Zhibin Li
Intrinsic Language-Guided Exploration for Complex Long-Horizon Robotic Manipulation Tasks
Accepted at the International Conference on Robotics and Automation (ICRA), 2024. The manuscript consists of 10 pages and 6 figures
null
null
null
cs.RO cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Current reinforcement learning algorithms struggle in sparse and complex environments, most notably in long-horizon manipulation tasks entailing a plethora of different sequences. In this work, we propose the Intrinsically Guided Exploration from Large Language Models (IGE-LLMs) framework. By leveraging LLMs as an assistive intrinsic reward, IGE-LLMs guides the exploratory process in reinforcement learning to address intricate long-horizon with sparse rewards robotic manipulation tasks. We evaluate our framework and related intrinsic learning methods in an environment challenged with exploration, and a complex robotic manipulation task challenged by both exploration and long-horizons. Results show IGE-LLMs (i) exhibit notably higher performance over related intrinsic methods and the direct use of LLMs in decision-making, (ii) can be combined and complement existing learning methods highlighting its modularity, (iii) are fairly insensitive to different intrinsic scaling parameters, and (iv) maintain robustness against increased levels of uncertainty and horizons.
[ { "created": "Thu, 28 Sep 2023 11:14:52 GMT", "version": "v1" }, { "created": "Thu, 7 Mar 2024 17:53:35 GMT", "version": "v2" } ]
2024-03-08
[ [ "Triantafyllidis", "Eleftherios", "" ], [ "Christianos", "Filippos", "" ], [ "Li", "Zhibin", "" ] ]
Current reinforcement learning algorithms struggle in sparse and complex environments, most notably in long-horizon manipulation tasks entailing a plethora of different sequences. In this work, we propose the Intrinsically Guided Exploration from Large Language Models (IGE-LLMs) framework. By leveraging LLMs as an assistive intrinsic reward, IGE-LLMs guides the exploratory process in reinforcement learning to address intricate long-horizon with sparse rewards robotic manipulation tasks. We evaluate our framework and related intrinsic learning methods in an environment challenged with exploration, and a complex robotic manipulation task challenged by both exploration and long-horizons. Results show IGE-LLMs (i) exhibit notably higher performance over related intrinsic methods and the direct use of LLMs in decision-making, (ii) can be combined and complement existing learning methods highlighting its modularity, (iii) are fairly insensitive to different intrinsic scaling parameters, and (iv) maintain robustness against increased levels of uncertainty and horizons.
2305.18445
Sunitha Basodi
Sunitha Basodi, Krishna Pusuluri, Xueli Xiao, Yi Pan
Intelligent gradient amplification for deep neural networks
null
null
null
null
cs.LG cs.CV
http://creativecommons.org/licenses/by-nc-nd/4.0/
Deep learning models offer superior performance compared to other machine learning techniques for a variety of tasks and domains, but pose their own challenges. In particular, deep learning models require larger training times as the depth of a model increases, and suffer from vanishing gradients. Several solutions address these problems independently, but there have been minimal efforts to identify an integrated solution that improves the performance of a model by addressing vanishing gradients, as well as accelerates the training process to achieve higher performance at larger learning rates. In this work, we intelligently determine which layers of a deep learning model to apply gradient amplification to, using a formulated approach that analyzes gradient fluctuations of layers during training. Detailed experiments are performed for simpler and deeper neural networks using two different intelligent measures and two different thresholds that determine the amplification layers, and a training strategy where gradients are amplified only during certain epochs. Results show that our amplification offers better performance compared to the original models, and achieves accuracy improvement of around 2.5% on CIFAR- 10 and around 4.5% on CIFAR-100 datasets, even when the models are trained with higher learning rates.
[ { "created": "Mon, 29 May 2023 03:38:09 GMT", "version": "v1" } ]
2023-05-31
[ [ "Basodi", "Sunitha", "" ], [ "Pusuluri", "Krishna", "" ], [ "Xiao", "Xueli", "" ], [ "Pan", "Yi", "" ] ]
Deep learning models offer superior performance compared to other machine learning techniques for a variety of tasks and domains, but pose their own challenges. In particular, deep learning models require larger training times as the depth of a model increases, and suffer from vanishing gradients. Several solutions address these problems independently, but there have been minimal efforts to identify an integrated solution that improves the performance of a model by addressing vanishing gradients, as well as accelerates the training process to achieve higher performance at larger learning rates. In this work, we intelligently determine which layers of a deep learning model to apply gradient amplification to, using a formulated approach that analyzes gradient fluctuations of layers during training. Detailed experiments are performed for simpler and deeper neural networks using two different intelligent measures and two different thresholds that determine the amplification layers, and a training strategy where gradients are amplified only during certain epochs. Results show that our amplification offers better performance compared to the original models, and achieves accuracy improvement of around 2.5% on CIFAR- 10 and around 4.5% on CIFAR-100 datasets, even when the models are trained with higher learning rates.
2001.02091
Chengkun Lang
Huiwei Zhou, Zhuang Liu1, Shixian Ning, Chengkun Lang, Yingyu Lin, Lei Du
Knowledge-aware Attention Network for Protein-Protein Interaction Extraction
Published on Journal of Biomedical Informatics, 14 pages, 5 figures
Journal of Biomedical Informatics, 2019, 96: 103234
10.1016/j.jbi.2019.103234
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Protein-protein interaction (PPI) extraction from published scientific literature provides additional support for precision medicine efforts. However, many of the current PPI extraction methods need extensive feature engineering and cannot make full use of the prior knowledge in knowledge bases (KB). KBs contain huge amounts of structured information about entities and relationships, therefore plays a pivotal role in PPI extraction. This paper proposes a knowledge-aware attention network (KAN) to fuse prior knowledge about protein-protein pairs and context information for PPI extraction. The proposed model first adopts a diagonal-disabled multi-head attention mechanism to encode context sequence along with knowledge representations learned from KB. Then a novel multi-dimensional attention mechanism is used to select the features that can best describe the encoded context. Experiment results on the BioCreative VI PPI dataset show that the proposed approach could acquire knowledge-aware dependencies between different words in a sequence and lead to a new state-of-the-art performance.
[ { "created": "Tue, 7 Jan 2020 15:02:28 GMT", "version": "v1" } ]
2020-01-08
[ [ "Zhou", "Huiwei", "" ], [ "Liu1", "Zhuang", "" ], [ "Ning", "Shixian", "" ], [ "Lang", "Chengkun", "" ], [ "Lin", "Yingyu", "" ], [ "Du", "Lei", "" ] ]
Protein-protein interaction (PPI) extraction from published scientific literature provides additional support for precision medicine efforts. However, many of the current PPI extraction methods need extensive feature engineering and cannot make full use of the prior knowledge in knowledge bases (KB). KBs contain huge amounts of structured information about entities and relationships, therefore plays a pivotal role in PPI extraction. This paper proposes a knowledge-aware attention network (KAN) to fuse prior knowledge about protein-protein pairs and context information for PPI extraction. The proposed model first adopts a diagonal-disabled multi-head attention mechanism to encode context sequence along with knowledge representations learned from KB. Then a novel multi-dimensional attention mechanism is used to select the features that can best describe the encoded context. Experiment results on the BioCreative VI PPI dataset show that the proposed approach could acquire knowledge-aware dependencies between different words in a sequence and lead to a new state-of-the-art performance.
1207.2567
Nadeem Javaid
S. Hayat, N. Javaid, Z. A. Khan, A. Shareef, A. Mahmood, S. H. Bouk
Energy Efficient MAC Protocols
null
5th AHPCN in conjunction with 14th HPCC-2012, Liverpool, UK
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents a survey of energy efficiency of Medium Access Control (MAC) protocols for Wireless Body Area Sensor Networks (WBASNs). We highlight the features of MAC protocols along with their advantages and limitations in context of WBASNs. Comparison of Low Power Listening (LPL), Scheduled Contention and Time Division Multiple Access (TDMA) is also elaborated. MAC protocols with respect to different approaches and techniques which are used for energy minimization, traffic control mechanisms for collision avoidance are discussed.We also present a survey of path loss models for In-body, On-body and Off-body communications in WBASNs and analytically discuss that path loss is maximum in In-body communication because of low energy levels to take care of tissues and organs located inside the body. Survey of Power model for WBANs of CSMA/CA and beacon mode is also presented.
[ { "created": "Wed, 11 Jul 2012 09:05:36 GMT", "version": "v1" } ]
2012-07-12
[ [ "Hayat", "S.", "" ], [ "Javaid", "N.", "" ], [ "Khan", "Z. A.", "" ], [ "Shareef", "A.", "" ], [ "Mahmood", "A.", "" ], [ "Bouk", "S. H.", "" ] ]
This paper presents a survey of energy efficiency of Medium Access Control (MAC) protocols for Wireless Body Area Sensor Networks (WBASNs). We highlight the features of MAC protocols along with their advantages and limitations in context of WBASNs. Comparison of Low Power Listening (LPL), Scheduled Contention and Time Division Multiple Access (TDMA) is also elaborated. MAC protocols with respect to different approaches and techniques which are used for energy minimization, traffic control mechanisms for collision avoidance are discussed.We also present a survey of path loss models for In-body, On-body and Off-body communications in WBASNs and analytically discuss that path loss is maximum in In-body communication because of low energy levels to take care of tissues and organs located inside the body. Survey of Power model for WBANs of CSMA/CA and beacon mode is also presented.
1910.06299
Gamal Sallam
Gamal Sallam, Zizhan Zheng, Bo Ji
Placement and Allocation of Virtual Network Functions: Multi-dimensional Case
null
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Network function virtualization (NFV) is an emerging design paradigm that replaces physical middlebox devices with software modules running on general purpose commodity servers. While gradually transitioning to NFV, Internet service providers face the problem of where to introduce NFV in order to make the most benefit of that; here, we measure the benefit by the amount of traffic that can be served in an NFV-enabled network. This problem is non-trivial as it is composed of two challenging subproblems: 1) placement of nodes to support virtual network functions (referred to as VNF-nodes); 2) allocation of the VNF-nodes' resources to network flows. This problem has been studied for the one-dimensional setting, where all network flows require one network function, which requires a unit of resource to process a unit of flow. In this work, we consider the multi-dimensional setting, where flows must be processed by multiple network functions, which require a different amount of each resource to process a unit of flow. The multi-dimensional setting introduces new challenges in addition to those of the one-dimensional setting (e.g., NP-hardness and non-submodularity) and also makes the resource allocation subproblem a multi-dimensional generalization of the generalized assignment problem with assignment restrictions. To address these difficulties, we propose a novel two-level relaxation method that allows us to draw a connection to the sequence submodular theory and utilize the property of sequence submodularity along with the primal-dual technique to design two approximation algorithms. We further prove that the proposed algorithms have a non-trivial approximation ratio that depends on the number of VNF-nodes, resources, and a measure of the available resource compared to flow demand. Finally, we perform trace-driven simulations to show the effectiveness of the proposed algorithms.
[ { "created": "Mon, 14 Oct 2019 17:27:59 GMT", "version": "v1" }, { "created": "Sun, 31 May 2020 22:39:43 GMT", "version": "v2" }, { "created": "Sat, 19 Feb 2022 18:41:15 GMT", "version": "v3" } ]
2022-02-22
[ [ "Sallam", "Gamal", "" ], [ "Zheng", "Zizhan", "" ], [ "Ji", "Bo", "" ] ]
Network function virtualization (NFV) is an emerging design paradigm that replaces physical middlebox devices with software modules running on general purpose commodity servers. While gradually transitioning to NFV, Internet service providers face the problem of where to introduce NFV in order to make the most benefit of that; here, we measure the benefit by the amount of traffic that can be served in an NFV-enabled network. This problem is non-trivial as it is composed of two challenging subproblems: 1) placement of nodes to support virtual network functions (referred to as VNF-nodes); 2) allocation of the VNF-nodes' resources to network flows. This problem has been studied for the one-dimensional setting, where all network flows require one network function, which requires a unit of resource to process a unit of flow. In this work, we consider the multi-dimensional setting, where flows must be processed by multiple network functions, which require a different amount of each resource to process a unit of flow. The multi-dimensional setting introduces new challenges in addition to those of the one-dimensional setting (e.g., NP-hardness and non-submodularity) and also makes the resource allocation subproblem a multi-dimensional generalization of the generalized assignment problem with assignment restrictions. To address these difficulties, we propose a novel two-level relaxation method that allows us to draw a connection to the sequence submodular theory and utilize the property of sequence submodularity along with the primal-dual technique to design two approximation algorithms. We further prove that the proposed algorithms have a non-trivial approximation ratio that depends on the number of VNF-nodes, resources, and a measure of the available resource compared to flow demand. Finally, we perform trace-driven simulations to show the effectiveness of the proposed algorithms.
2005.13751
Iraklis Moutidis
Iraklis Moutidis and Hywel T.P. Williams
Complex networks for event detection in heterogeneous high volume news streams
null
null
null
null
cs.SI cs.CL cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Detecting important events in high volume news streams is an important task for a variety of purposes.The volume and rate of online news increases the need for automated event detection methods thatcan operate in real time. In this paper we develop a network-based approach that makes the workingassumption that important news events always involve named entities (such as persons, locationsand organizations) that are linked in news articles. Our approach uses natural language processingtechniques to detect these entities in a stream of news articles and then creates a time-stamped seriesof networks in which the detected entities are linked by co-occurrence in articles and sentences. Inthis prototype, weighted node degree is tracked over time and change-point detection used to locateimportant events. Potential events are characterized and distinguished using community detectionon KeyGraphs that relate named entities and informative noun-phrases from related articles. Thismethodology already produces promising results and will be extended in future to include a widervariety of complex network analysis techniques.
[ { "created": "Thu, 28 May 2020 02:45:43 GMT", "version": "v1" } ]
2020-05-29
[ [ "Moutidis", "Iraklis", "" ], [ "Williams", "Hywel T. P.", "" ] ]
Detecting important events in high volume news streams is an important task for a variety of purposes.The volume and rate of online news increases the need for automated event detection methods thatcan operate in real time. In this paper we develop a network-based approach that makes the workingassumption that important news events always involve named entities (such as persons, locationsand organizations) that are linked in news articles. Our approach uses natural language processingtechniques to detect these entities in a stream of news articles and then creates a time-stamped seriesof networks in which the detected entities are linked by co-occurrence in articles and sentences. Inthis prototype, weighted node degree is tracked over time and change-point detection used to locateimportant events. Potential events are characterized and distinguished using community detectionon KeyGraphs that relate named entities and informative noun-phrases from related articles. Thismethodology already produces promising results and will be extended in future to include a widervariety of complex network analysis techniques.
2209.15351
Xiuzhen Guo
Xiuzhen Guo, Longfei Shangguan, Yuan He, Jia Zhang, Haotian Jiang, Awais Ahmad Siddiqi, Yunhao Liu
Efficient Ambient LoRa Backscatter with On-Off Keying Modulation
null
null
null
null
cs.NI eess.SP
http://creativecommons.org/licenses/by/4.0/
Backscatter communication holds potential for ubiquitous and low-cost connectivity among low-power IoT devices. To avoid interference between the carrier signal and the backscatter signal, recent works propose a frequency-shifting technique to separate these two signals in the frequency domain. Such proposals, however, have to occupy the precious wireless spectrum that is already overcrowded, and increase the power, cost, and complexity of the backscatter tag. In this paper, we revisit the classic ON-OFF Keying (OOK) modulation and propose Aloba, a backscatter system that takes the ambient LoRa transmissions as the excitation and piggybacks the in-band OOK modulated signals over the LoRa transmissions. Our design enables the backsactter signal to work in the same frequency band of the carrier signal, meanwhile achieving flexible data rate at different transmission range. The key contributions of Aloba include: (1) the design of a low-power backscatter tag that can pick up the ambient LoRa signals from other signals. (2) a novel decoding algorithm to demodulate both the carrier signal and the backscatter signal from their superposition. We further adopt link coding mechanism and interleave operation to enhance the reliability of backscatter signal decoding. We implement Aloba and conduct head-to-head comparison with the state-of-the-art LoRa backscatter system PLoRa in various settings. The experiment results show Aloba can achieve 199.4 Kbps data rate at various distances, 52.4 times higher than PLoRa.
[ { "created": "Fri, 30 Sep 2022 10:16:43 GMT", "version": "v1" } ]
2022-10-03
[ [ "Guo", "Xiuzhen", "" ], [ "Shangguan", "Longfei", "" ], [ "He", "Yuan", "" ], [ "Zhang", "Jia", "" ], [ "Jiang", "Haotian", "" ], [ "Siddiqi", "Awais Ahmad", "" ], [ "Liu", "Yunhao", "" ] ]
Backscatter communication holds potential for ubiquitous and low-cost connectivity among low-power IoT devices. To avoid interference between the carrier signal and the backscatter signal, recent works propose a frequency-shifting technique to separate these two signals in the frequency domain. Such proposals, however, have to occupy the precious wireless spectrum that is already overcrowded, and increase the power, cost, and complexity of the backscatter tag. In this paper, we revisit the classic ON-OFF Keying (OOK) modulation and propose Aloba, a backscatter system that takes the ambient LoRa transmissions as the excitation and piggybacks the in-band OOK modulated signals over the LoRa transmissions. Our design enables the backsactter signal to work in the same frequency band of the carrier signal, meanwhile achieving flexible data rate at different transmission range. The key contributions of Aloba include: (1) the design of a low-power backscatter tag that can pick up the ambient LoRa signals from other signals. (2) a novel decoding algorithm to demodulate both the carrier signal and the backscatter signal from their superposition. We further adopt link coding mechanism and interleave operation to enhance the reliability of backscatter signal decoding. We implement Aloba and conduct head-to-head comparison with the state-of-the-art LoRa backscatter system PLoRa in various settings. The experiment results show Aloba can achieve 199.4 Kbps data rate at various distances, 52.4 times higher than PLoRa.
1808.01688
Huan Zhang
Dong Su, Huan Zhang, Hongge Chen, Jinfeng Yi, Pin-Yu Chen, Yupeng Gao
Is Robustness the Cost of Accuracy? -- A Comprehensive Study on the Robustness of 18 Deep Image Classification Models
Accepted by the European Conference on Computer Vision (ECCV) 2018
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The prediction accuracy has been the long-lasting and sole standard for comparing the performance of different image classification models, including the ImageNet competition. However, recent studies have highlighted the lack of robustness in well-trained deep neural networks to adversarial examples. Visually imperceptible perturbations to natural images can easily be crafted and mislead the image classifiers towards misclassification. To demystify the trade-offs between robustness and accuracy, in this paper we thoroughly benchmark 18 ImageNet models using multiple robustness metrics, including the distortion, success rate and transferability of adversarial examples between 306 pairs of models. Our extensive experimental results reveal several new insights: (1) linear scaling law - the empirical $\ell_2$ and $\ell_\infty$ distortion metrics scale linearly with the logarithm of classification error; (2) model architecture is a more critical factor to robustness than model size, and the disclosed accuracy-robustness Pareto frontier can be used as an evaluation criterion for ImageNet model designers; (3) for a similar network architecture, increasing network depth slightly improves robustness in $\ell_\infty$ distortion; (4) there exist models (in VGG family) that exhibit high adversarial transferability, while most adversarial examples crafted from one model can only be transferred within the same family. Experiment code is publicly available at \url{https://github.com/huanzhang12/Adversarial_Survey}.
[ { "created": "Sun, 5 Aug 2018 21:43:01 GMT", "version": "v1" }, { "created": "Mon, 4 Mar 2019 00:35:26 GMT", "version": "v2" } ]
2019-03-05
[ [ "Su", "Dong", "" ], [ "Zhang", "Huan", "" ], [ "Chen", "Hongge", "" ], [ "Yi", "Jinfeng", "" ], [ "Chen", "Pin-Yu", "" ], [ "Gao", "Yupeng", "" ] ]
The prediction accuracy has been the long-lasting and sole standard for comparing the performance of different image classification models, including the ImageNet competition. However, recent studies have highlighted the lack of robustness in well-trained deep neural networks to adversarial examples. Visually imperceptible perturbations to natural images can easily be crafted and mislead the image classifiers towards misclassification. To demystify the trade-offs between robustness and accuracy, in this paper we thoroughly benchmark 18 ImageNet models using multiple robustness metrics, including the distortion, success rate and transferability of adversarial examples between 306 pairs of models. Our extensive experimental results reveal several new insights: (1) linear scaling law - the empirical $\ell_2$ and $\ell_\infty$ distortion metrics scale linearly with the logarithm of classification error; (2) model architecture is a more critical factor to robustness than model size, and the disclosed accuracy-robustness Pareto frontier can be used as an evaluation criterion for ImageNet model designers; (3) for a similar network architecture, increasing network depth slightly improves robustness in $\ell_\infty$ distortion; (4) there exist models (in VGG family) that exhibit high adversarial transferability, while most adversarial examples crafted from one model can only be transferred within the same family. Experiment code is publicly available at \url{https://github.com/huanzhang12/Adversarial_Survey}.
2303.10087
Shuai Chen
Shuai Chen, Yash Bhalgat, Xinghui Li, Jiawang Bian, Kejie Li, Zirui Wang, Victor Adrian Prisacariu
Neural Refinement for Absolute Pose Regression with Feature Synthesis
Paper Accepted by CVPR 2024. Project Page: http://nefes.active.vision. Code will be released at https://github.com/ActiveVisionLab/NeFeS
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Absolute Pose Regression (APR) methods use deep neural networks to directly regress camera poses from RGB images. However, the predominant APR architectures only rely on 2D operations during inference, resulting in limited accuracy of pose estimation due to the lack of 3D geometry constraints or priors. In this work, we propose a test-time refinement pipeline that leverages implicit geometric constraints using a robust feature field to enhance the ability of APR methods to use 3D information during inference. We also introduce a novel Neural Feature Synthesizer (NeFeS) model, which encodes 3D geometric features during training and directly renders dense novel view features at test time to refine APR methods. To enhance the robustness of our model, we introduce a feature fusion module and a progressive training strategy. Our proposed method achieves state-of-the-art single-image APR accuracy on indoor and outdoor datasets.
[ { "created": "Fri, 17 Mar 2023 16:10:50 GMT", "version": "v1" }, { "created": "Fri, 1 Mar 2024 01:40:52 GMT", "version": "v2" } ]
2024-03-04
[ [ "Chen", "Shuai", "" ], [ "Bhalgat", "Yash", "" ], [ "Li", "Xinghui", "" ], [ "Bian", "Jiawang", "" ], [ "Li", "Kejie", "" ], [ "Wang", "Zirui", "" ], [ "Prisacariu", "Victor Adrian", "" ] ]
Absolute Pose Regression (APR) methods use deep neural networks to directly regress camera poses from RGB images. However, the predominant APR architectures only rely on 2D operations during inference, resulting in limited accuracy of pose estimation due to the lack of 3D geometry constraints or priors. In this work, we propose a test-time refinement pipeline that leverages implicit geometric constraints using a robust feature field to enhance the ability of APR methods to use 3D information during inference. We also introduce a novel Neural Feature Synthesizer (NeFeS) model, which encodes 3D geometric features during training and directly renders dense novel view features at test time to refine APR methods. To enhance the robustness of our model, we introduce a feature fusion module and a progressive training strategy. Our proposed method achieves state-of-the-art single-image APR accuracy on indoor and outdoor datasets.
2306.11112
Emily Diana
Emily Diana and Alexander Williams Tolbert
Correcting Underrepresentation and Intersectional Bias for Classification
null
null
null
null
cs.LG cs.CY cs.DS stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the problem of learning from data corrupted by underrepresentation bias, where positive examples are filtered from the data at different, unknown rates for a fixed number of sensitive groups. We show that with a small amount of unbiased data, we can efficiently estimate the group-wise drop-out rates, even in settings where intersectional group membership makes learning each intersectional rate computationally infeasible. Using these estimates, we construct a reweighting scheme that allows us to approximate the loss of any hypothesis on the true distribution, even if we only observe the empirical error on a biased sample. From this, we present an algorithm encapsulating this learning and reweighting process along with a thorough empirical investigation. Finally, we define a bespoke notion of PAC learnability for the underrepresentation and intersectional bias setting and show that our algorithm permits efficient learning for model classes of finite VC dimension.
[ { "created": "Mon, 19 Jun 2023 18:25:44 GMT", "version": "v1" }, { "created": "Wed, 19 Jul 2023 21:08:41 GMT", "version": "v2" }, { "created": "Tue, 26 Sep 2023 19:02:34 GMT", "version": "v3" }, { "created": "Mon, 3 Jun 2024 20:57:56 GMT", "version": "v4" } ]
2024-06-05
[ [ "Diana", "Emily", "" ], [ "Tolbert", "Alexander Williams", "" ] ]
We consider the problem of learning from data corrupted by underrepresentation bias, where positive examples are filtered from the data at different, unknown rates for a fixed number of sensitive groups. We show that with a small amount of unbiased data, we can efficiently estimate the group-wise drop-out rates, even in settings where intersectional group membership makes learning each intersectional rate computationally infeasible. Using these estimates, we construct a reweighting scheme that allows us to approximate the loss of any hypothesis on the true distribution, even if we only observe the empirical error on a biased sample. From this, we present an algorithm encapsulating this learning and reweighting process along with a thorough empirical investigation. Finally, we define a bespoke notion of PAC learnability for the underrepresentation and intersectional bias setting and show that our algorithm permits efficient learning for model classes of finite VC dimension.
2108.04343
Maryem Rhanoui
Siham Yousfi and Maryem Rhanoui and Dalila Chiadmi
Towards a Generic Multimodal Architecture for Batch and Streaming Big Data Integration
null
Journal of Computer Science, Volume 15 No. 1, 2019, 207-220
10.3844/jcssp.2019.207.220
null
cs.AI cs.LG
http://creativecommons.org/licenses/by-nc-sa/4.0/
Big Data are rapidly produced from various heterogeneous data sources. They are of different types (text, image, video or audio) and have different levels of reliability and completeness. One of the most interesting architectures that deal with the large amount of emerging data at high velocity is called the lambda architecture. In fact, it combines two different processing layers namely batch and speed layers, each providing specific views of data while ensuring robustness, fast and scalable data processing. However, most papers dealing with the lambda architecture are focusing one single type of data generally produced by a single data source. Besides, the layers of the architecture are implemented independently, or, at best, are combined to perform basic processing without assessing either the data reliability or completeness. Therefore, inspired by the lambda architecture, we propose in this paper a generic multimodal architecture that combines both batch and streaming processing in order to build a complete, global and accurate insight in near-real-time based on the knowledge extracted from multiple heterogeneous Big Data sources. Our architecture uses batch processing to analyze the data structures and contents, build the learning models and calculate the reliability index of the involved sources, while the streaming processing uses the built-in models of the batch layer to immediately process incoming data and rapidly provide results. We validate our architecture in the context of urban traffic management systems in order to detect congestions.
[ { "created": "Mon, 9 Aug 2021 20:50:01 GMT", "version": "v1" } ]
2021-08-11
[ [ "Yousfi", "Siham", "" ], [ "Rhanoui", "Maryem", "" ], [ "Chiadmi", "Dalila", "" ] ]
Big Data are rapidly produced from various heterogeneous data sources. They are of different types (text, image, video or audio) and have different levels of reliability and completeness. One of the most interesting architectures that deal with the large amount of emerging data at high velocity is called the lambda architecture. In fact, it combines two different processing layers namely batch and speed layers, each providing specific views of data while ensuring robustness, fast and scalable data processing. However, most papers dealing with the lambda architecture are focusing one single type of data generally produced by a single data source. Besides, the layers of the architecture are implemented independently, or, at best, are combined to perform basic processing without assessing either the data reliability or completeness. Therefore, inspired by the lambda architecture, we propose in this paper a generic multimodal architecture that combines both batch and streaming processing in order to build a complete, global and accurate insight in near-real-time based on the knowledge extracted from multiple heterogeneous Big Data sources. Our architecture uses batch processing to analyze the data structures and contents, build the learning models and calculate the reliability index of the involved sources, while the streaming processing uses the built-in models of the batch layer to immediately process incoming data and rapidly provide results. We validate our architecture in the context of urban traffic management systems in order to detect congestions.
0810.1424
Sidharth Jaggi
Bikash Kumar Dey, Sidharth Jaggi, and Michael Langberg
"Real" Slepian-Wolf Codes
20 pages. Preliminary version presented at ISIT 2008, Toronto, Canada
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We provide a novel achievability proof of the Slepian-Wolf theorem for i.i.d. sources over finite alphabets. We demonstrate that random codes that are linear over the real field achieve the classical Slepian-Wolf rate-region. For finite alphabets we show that typicality decoding is equivalent to solving an integer program. Minimum entropy decoding is also shown to achieve exponentially small probability of error. The techniques used may be of independent interest for code design for a wide class of information theory problems, and for the field of compressed sensing.
[ { "created": "Wed, 8 Oct 2008 12:57:51 GMT", "version": "v1" } ]
2008-10-09
[ [ "Dey", "Bikash Kumar", "" ], [ "Jaggi", "Sidharth", "" ], [ "Langberg", "Michael", "" ] ]
We provide a novel achievability proof of the Slepian-Wolf theorem for i.i.d. sources over finite alphabets. We demonstrate that random codes that are linear over the real field achieve the classical Slepian-Wolf rate-region. For finite alphabets we show that typicality decoding is equivalent to solving an integer program. Minimum entropy decoding is also shown to achieve exponentially small probability of error. The techniques used may be of independent interest for code design for a wide class of information theory problems, and for the field of compressed sensing.
2204.05311
M.Z. Naser
M.Z. Naser, Aybike Ozyuksel Ciftcioglu
Causal Discovery and Causal Learning for Fire Resistance Evaluation: Incorporating Domain Knowledge
null
null
null
null
cs.LG stat.AP
http://creativecommons.org/licenses/by-nc-nd/4.0/
Experiments remain the gold standard to establish an understanding of fire-related phenomena. A primary goal in designing tests is to uncover the data generating process (i.e., the how and why the observations we see come to be); or simply what causes such observations. Uncovering such a process not only advances our knowledge but also provides us with the capability to be able to predict phenomena accurately. This paper presents an approach that leverages causal discovery and causal inference to evaluate the fire resistance of structural members. In this approach, causal discovery algorithms are adopted to uncover the causal structure between key variables pertaining to the fire resistance of reinforced concrete (RC) columns. Then, companion inference algorithms are applied to infer (estimate) the influence of each variable on the fire resistance given a specific intervention. Finally, this study ends by contrasting the algorithmic causal discovery with that obtained from domain knowledge and traditional machine learning. Our findings clearly show the potential and merit of adopting causality into our domain.
[ { "created": "Mon, 11 Apr 2022 22:35:52 GMT", "version": "v1" } ]
2022-04-13
[ [ "Naser", "M. Z.", "" ], [ "Ciftcioglu", "Aybike Ozyuksel", "" ] ]
Experiments remain the gold standard to establish an understanding of fire-related phenomena. A primary goal in designing tests is to uncover the data generating process (i.e., the how and why the observations we see come to be); or simply what causes such observations. Uncovering such a process not only advances our knowledge but also provides us with the capability to be able to predict phenomena accurately. This paper presents an approach that leverages causal discovery and causal inference to evaluate the fire resistance of structural members. In this approach, causal discovery algorithms are adopted to uncover the causal structure between key variables pertaining to the fire resistance of reinforced concrete (RC) columns. Then, companion inference algorithms are applied to infer (estimate) the influence of each variable on the fire resistance given a specific intervention. Finally, this study ends by contrasting the algorithmic causal discovery with that obtained from domain knowledge and traditional machine learning. Our findings clearly show the potential and merit of adopting causality into our domain.
2211.05295
Jianye Yi
Jianye Yi, Xiaopin Zhong, Weixiang Liu, Zongze Wu, Yuanlong Deng and Zhengguang Wu
Harmonizing output imbalance for defect segmentation on extremely-imbalanced photovoltaic module cells images
19 pages, 16 figures, 3 appendixes
null
null
null
cs.CV cs.LG eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The continuous development of the photovoltaic (PV) industry has raised high requirements for the quality of monocrystalline of PV module cells. When learning to segment defect regions in PV module cell images, Tiny Hidden Cracks (THC) lead to extremely-imbalanced samples. The ratio of defect pixels to normal pixels can be as low as 1:2000. This extreme imbalance makes it difficult to segment the THC of PV module cells, which is also a challenge for semantic segmentation. To address the problem of segmenting defects on extremely-imbalanced THC data, the paper makes contributions from three aspects: (1) it proposes an explicit measure for output imbalance; (2) it generalizes a distribution-based loss that can handle different types of output imbalances; and (3) it introduces a compound loss with our adaptive hyperparameter selection algorithm that can keep the consistency of training and inference for harmonizing the output imbalance on extremelyimbalanced input data. The proposed method is evaluated on four widely-used deep learning architectures and four datasets with varying degrees of input imbalance. The experimental results show that the proposed method outperforms existing methods.
[ { "created": "Thu, 10 Nov 2022 02:05:17 GMT", "version": "v1" }, { "created": "Mon, 12 Dec 2022 04:18:22 GMT", "version": "v2" }, { "created": "Tue, 13 Dec 2022 02:09:40 GMT", "version": "v3" }, { "created": "Tue, 24 Oct 2023 10:08:10 GMT", "version": "v4" } ]
2023-10-25
[ [ "Yi", "Jianye", "" ], [ "Zhong", "Xiaopin", "" ], [ "Liu", "Weixiang", "" ], [ "Wu", "Zongze", "" ], [ "Deng", "Yuanlong", "" ], [ "Wu", "Zhengguang", "" ] ]
The continuous development of the photovoltaic (PV) industry has raised high requirements for the quality of monocrystalline of PV module cells. When learning to segment defect regions in PV module cell images, Tiny Hidden Cracks (THC) lead to extremely-imbalanced samples. The ratio of defect pixels to normal pixels can be as low as 1:2000. This extreme imbalance makes it difficult to segment the THC of PV module cells, which is also a challenge for semantic segmentation. To address the problem of segmenting defects on extremely-imbalanced THC data, the paper makes contributions from three aspects: (1) it proposes an explicit measure for output imbalance; (2) it generalizes a distribution-based loss that can handle different types of output imbalances; and (3) it introduces a compound loss with our adaptive hyperparameter selection algorithm that can keep the consistency of training and inference for harmonizing the output imbalance on extremelyimbalanced input data. The proposed method is evaluated on four widely-used deep learning architectures and four datasets with varying degrees of input imbalance. The experimental results show that the proposed method outperforms existing methods.
2105.05716
Adrian Remonda
Adrian Remonda, Eduardo Veas, Granit Luzhnica
Acting upon Imagination: when to trust imagined trajectories in model based reinforcement learning
null
null
null
null
cs.AI cs.LG
http://creativecommons.org/licenses/by/4.0/
Model-based reinforcement learning (MBRL) aims to learn model(s) of the environment dynamics that can predict the outcome of its actions. Forward application of the model yields so called imagined trajectories (sequences of action, predicted state-reward) used to optimize the set of candidate actions that maximize expected reward. The outcome, an ideal imagined trajectory or plan, is imperfect and typically MBRL relies on model predictive control (MPC) to overcome this by continuously re-planning from scratch, incurring thus major computational cost and increasing complexity in tasks with longer receding horizon. We propose uncertainty estimation methods for online evaluation of imagined trajectories to assess whether further planned actions can be trusted to deliver acceptable reward. These methods include comparing the error after performing the last action with the standard expected error and using model uncertainty to assess the deviation from expected outcomes. Additionally, we introduce methods that exploit the forward propagation of the dynamics model to evaluate if the remainder of the plan aligns with expected results and assess the remainder of the plan in terms of the expected reward. Our experiments demonstrate the effectiveness of the proposed uncertainty estimation methods by applying them to avoid unnecessary trajectory replanning in a shooting MBRL setting. Results highlight significant reduction on computational costs without sacrificing performance.
[ { "created": "Wed, 12 May 2021 15:04:07 GMT", "version": "v1" }, { "created": "Thu, 13 May 2021 10:26:30 GMT", "version": "v2" }, { "created": "Wed, 9 Nov 2022 17:34:08 GMT", "version": "v3" }, { "created": "Fri, 11 Nov 2022 10:47:43 GMT", "version": "v4" }, { "created": "Thu, 18 Apr 2024 23:45:00 GMT", "version": "v5" }, { "created": "Tue, 30 Jul 2024 14:25:07 GMT", "version": "v6" } ]
2024-07-31
[ [ "Remonda", "Adrian", "" ], [ "Veas", "Eduardo", "" ], [ "Luzhnica", "Granit", "" ] ]
Model-based reinforcement learning (MBRL) aims to learn model(s) of the environment dynamics that can predict the outcome of its actions. Forward application of the model yields so called imagined trajectories (sequences of action, predicted state-reward) used to optimize the set of candidate actions that maximize expected reward. The outcome, an ideal imagined trajectory or plan, is imperfect and typically MBRL relies on model predictive control (MPC) to overcome this by continuously re-planning from scratch, incurring thus major computational cost and increasing complexity in tasks with longer receding horizon. We propose uncertainty estimation methods for online evaluation of imagined trajectories to assess whether further planned actions can be trusted to deliver acceptable reward. These methods include comparing the error after performing the last action with the standard expected error and using model uncertainty to assess the deviation from expected outcomes. Additionally, we introduce methods that exploit the forward propagation of the dynamics model to evaluate if the remainder of the plan aligns with expected results and assess the remainder of the plan in terms of the expected reward. Our experiments demonstrate the effectiveness of the proposed uncertainty estimation methods by applying them to avoid unnecessary trajectory replanning in a shooting MBRL setting. Results highlight significant reduction on computational costs without sacrificing performance.
2309.08369
Yinqi Li
Zhupeng Ye, Yinqi Li, Zejian Yuan
An Efficient Wide-Range Pseudo-3D Vehicle Detection Using A Single Camera
11 pages, 27 figures
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Wide-range and fine-grained vehicle detection plays a critical role in enabling active safety features in intelligent driving systems. However, existing vehicle detection methods based on rectangular bounding boxes (BBox) often struggle with perceiving wide-range objects, especially small objects at long distances. And BBox expression cannot provide detailed geometric shape and pose information of vehicles. This paper proposes a novel wide-range Pseudo-3D Vehicle Detection method based on images from a single camera and incorporates efficient learning methods. This model takes a spliced image as input, which is obtained by combining two sub-window images from a high-resolution image. This image format maximizes the utilization of limited image resolution to retain essential information about wide-range vehicle objects. To detect pseudo-3D objects, our model adopts specifically designed detection heads. These heads simultaneously output extended BBox and Side Projection Line (SPL) representations, which capture vehicle shapes and poses, enabling high-precision detection. To further enhance the performance of detection, a joint constraint loss combining both the object box and SPL is designed during model training, improving the efficiency, stability, and prediction accuracy of the model. Experimental results on our self-built dataset demonstrate that our model achieves favorable performance in wide-range pseudo-3D vehicle detection across multiple evaluation metrics. Our demo video has been placed at https://www.youtube.com/watch?v=1gk1PmsQ5Q8.
[ { "created": "Fri, 15 Sep 2023 12:50:09 GMT", "version": "v1" } ]
2023-09-18
[ [ "Ye", "Zhupeng", "" ], [ "Li", "Yinqi", "" ], [ "Yuan", "Zejian", "" ] ]
Wide-range and fine-grained vehicle detection plays a critical role in enabling active safety features in intelligent driving systems. However, existing vehicle detection methods based on rectangular bounding boxes (BBox) often struggle with perceiving wide-range objects, especially small objects at long distances. And BBox expression cannot provide detailed geometric shape and pose information of vehicles. This paper proposes a novel wide-range Pseudo-3D Vehicle Detection method based on images from a single camera and incorporates efficient learning methods. This model takes a spliced image as input, which is obtained by combining two sub-window images from a high-resolution image. This image format maximizes the utilization of limited image resolution to retain essential information about wide-range vehicle objects. To detect pseudo-3D objects, our model adopts specifically designed detection heads. These heads simultaneously output extended BBox and Side Projection Line (SPL) representations, which capture vehicle shapes and poses, enabling high-precision detection. To further enhance the performance of detection, a joint constraint loss combining both the object box and SPL is designed during model training, improving the efficiency, stability, and prediction accuracy of the model. Experimental results on our self-built dataset demonstrate that our model achieves favorable performance in wide-range pseudo-3D vehicle detection across multiple evaluation metrics. Our demo video has been placed at https://www.youtube.com/watch?v=1gk1PmsQ5Q8.
1805.01129
Palakorn Achananuparp
Palakorn Achananuparp, Ee-Peng Lim, Vibhanshu Abhishek
Does Journaling Encourage Healthier Choices? Analyzing Healthy Eating Behaviors of Food Journalers
Published at Digital Health 2018
null
10.1145/3194658.3194663
null
cs.SI cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Past research has shown the benefits of food journaling in promoting mindful eating and healthier food choices. However, the links between journaling and healthy eating have not been thoroughly examined. Beyond caloric restriction, do journalers consistently and sufficiently consume healthful diets? How different are their eating habits compared to those of average consumers who tend to be less conscious about health? In this study, we analyze the healthy eating behaviors of active food journalers using data from MyFitnessPal. Surprisingly, our findings show that food journalers do not eat as healthily as they should despite their proclivity to health eating and their food choices resemble those of the general populace. Furthermore, we find that the journaling duration is only a marginal determinant of healthy eating outcomes and sociodemographic factors, such as gender and regions of residence, are much more predictive of healthy food choices.
[ { "created": "Thu, 3 May 2018 05:59:22 GMT", "version": "v1" } ]
2020-10-28
[ [ "Achananuparp", "Palakorn", "" ], [ "Lim", "Ee-Peng", "" ], [ "Abhishek", "Vibhanshu", "" ] ]
Past research has shown the benefits of food journaling in promoting mindful eating and healthier food choices. However, the links between journaling and healthy eating have not been thoroughly examined. Beyond caloric restriction, do journalers consistently and sufficiently consume healthful diets? How different are their eating habits compared to those of average consumers who tend to be less conscious about health? In this study, we analyze the healthy eating behaviors of active food journalers using data from MyFitnessPal. Surprisingly, our findings show that food journalers do not eat as healthily as they should despite their proclivity to health eating and their food choices resemble those of the general populace. Furthermore, we find that the journaling duration is only a marginal determinant of healthy eating outcomes and sociodemographic factors, such as gender and regions of residence, are much more predictive of healthy food choices.
2310.02076
Cindy Xiong Bearfield
Cindy Xiong Bearfield, Chase Stokes, Andrew Lovett, Steven Franconeri
What Does the Chart Say? Grouping Cues Guide Viewer Comparisons and Conclusions in Bar Charts
null
null
10.1109/TVCG.2023.3289292
null
cs.HC
http://creativecommons.org/licenses/by-nc-nd/4.0/
Reading a visualization is like reading a paragraph. Each sentence is a comparison: the mean of these is higher than those; this difference is smaller than that. What determines which comparisons are made first? The viewer's goals and expertise matter, but the way that values are visually grouped together within the chart also impacts those comparisons. Research from psychology suggests that comparisons involve multiple steps. First, the viewer divides the visualization into a set of units. This might include a single bar or a grouped set of bars. Then the viewer selects and compares two of these units, perhaps noting that one pair of bars is longer than another. Viewers might take an additional third step and perform a second-order comparison, perhaps determining that the difference between one pair of bars is greater than the difference between another pair. We create a visual comparison taxonomy that allows us to develop and test a sequence of hypotheses about which comparisons people are more likely to make when reading a visualization. We find that people tend to compare two groups before comparing two individual bars and that second-order comparisons are rare. Visual cues like spatial proximity and color can influence which elements are grouped together and selected for comparison, with spatial proximity being a stronger grouping cue. Interestingly, once the viewer grouped together and compared a set of bars, regardless of whether the group is formed by spatial proximity or color similarity, they no longer consider other possible groupings in their comparisons.
[ { "created": "Tue, 3 Oct 2023 14:16:25 GMT", "version": "v1" } ]
2023-10-04
[ [ "Bearfield", "Cindy Xiong", "" ], [ "Stokes", "Chase", "" ], [ "Lovett", "Andrew", "" ], [ "Franconeri", "Steven", "" ] ]
Reading a visualization is like reading a paragraph. Each sentence is a comparison: the mean of these is higher than those; this difference is smaller than that. What determines which comparisons are made first? The viewer's goals and expertise matter, but the way that values are visually grouped together within the chart also impacts those comparisons. Research from psychology suggests that comparisons involve multiple steps. First, the viewer divides the visualization into a set of units. This might include a single bar or a grouped set of bars. Then the viewer selects and compares two of these units, perhaps noting that one pair of bars is longer than another. Viewers might take an additional third step and perform a second-order comparison, perhaps determining that the difference between one pair of bars is greater than the difference between another pair. We create a visual comparison taxonomy that allows us to develop and test a sequence of hypotheses about which comparisons people are more likely to make when reading a visualization. We find that people tend to compare two groups before comparing two individual bars and that second-order comparisons are rare. Visual cues like spatial proximity and color can influence which elements are grouped together and selected for comparison, with spatial proximity being a stronger grouping cue. Interestingly, once the viewer grouped together and compared a set of bars, regardless of whether the group is formed by spatial proximity or color similarity, they no longer consider other possible groupings in their comparisons.
2404.15697
Orazio Pontorno
Orazio Pontorno (1), Luca Guarnera (1), Sebastiano Battiato (1) ((1) University of Catania)
DeepFeatureX Net: Deep Features eXtractors based Network for discriminating synthetic from real images
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deepfakes, synthetic images generated by deep learning algorithms, represent one of the biggest challenges in the field of Digital Forensics. The scientific community is working to develop approaches that can discriminate the origin of digital images (real or AI-generated). However, these methodologies face the challenge of generalization, that is, the ability to discern the nature of an image even if it is generated by an architecture not seen during training. This usually leads to a drop in performance. In this context, we propose a novel approach based on three blocks called Base Models, each of which is responsible for extracting the discriminative features of a specific image class (Diffusion Model-generated, GAN-generated, or real) as it is trained by exploiting deliberately unbalanced datasets. The features extracted from each block are then concatenated and processed to discriminate the origin of the input image. Experimental results showed that this approach not only demonstrates good robust capabilities to JPEG compression but also outperforms state-of-the-art methods in several generalization tests. Code, models and dataset are available at https://github.com/opontorno/block-based_deepfake-detection.
[ { "created": "Wed, 24 Apr 2024 07:25:36 GMT", "version": "v1" } ]
2024-04-25
[ [ "Pontorno", "Orazio", "" ], [ "Guarnera", "Luca", "" ], [ "Battiato", "Sebastiano", "" ] ]
Deepfakes, synthetic images generated by deep learning algorithms, represent one of the biggest challenges in the field of Digital Forensics. The scientific community is working to develop approaches that can discriminate the origin of digital images (real or AI-generated). However, these methodologies face the challenge of generalization, that is, the ability to discern the nature of an image even if it is generated by an architecture not seen during training. This usually leads to a drop in performance. In this context, we propose a novel approach based on three blocks called Base Models, each of which is responsible for extracting the discriminative features of a specific image class (Diffusion Model-generated, GAN-generated, or real) as it is trained by exploiting deliberately unbalanced datasets. The features extracted from each block are then concatenated and processed to discriminate the origin of the input image. Experimental results showed that this approach not only demonstrates good robust capabilities to JPEG compression but also outperforms state-of-the-art methods in several generalization tests. Code, models and dataset are available at https://github.com/opontorno/block-based_deepfake-detection.
1311.0413
Gordana Dodig Crnkovic
Gordana Dodig-Crnkovic
Information, Computation, Cognition. Agency-based Hierarchies of Levels
5 pages
null
null
null
cs.AI
http://creativecommons.org/licenses/by-nc-sa/3.0/
Nature can be seen as informational structure with computational dynamics (info-computationalism), where an (info-computational) agent is needed for the potential information of the world to actualize. Starting from the definition of information as the difference in one physical system that makes a difference in another physical system, which combines Bateson and Hewitt definitions, the argument is advanced for natural computation as a computational model of the dynamics of the physical world where information processing is constantly going on, on a variety of levels of organization. This setting helps elucidating the relationships between computation, information, agency and cognition, within the common conceptual framework, which has special relevance for biology and robotics.
[ { "created": "Sat, 2 Nov 2013 21:33:11 GMT", "version": "v1" } ]
2013-11-05
[ [ "Dodig-Crnkovic", "Gordana", "" ] ]
Nature can be seen as informational structure with computational dynamics (info-computationalism), where an (info-computational) agent is needed for the potential information of the world to actualize. Starting from the definition of information as the difference in one physical system that makes a difference in another physical system, which combines Bateson and Hewitt definitions, the argument is advanced for natural computation as a computational model of the dynamics of the physical world where information processing is constantly going on, on a variety of levels of organization. This setting helps elucidating the relationships between computation, information, agency and cognition, within the common conceptual framework, which has special relevance for biology and robotics.
2306.05582
Denizhan Oak
Denizhan Pak, Donsuk Lee, Samantha M. W. Wood, Justin N. Wood
A newborn embodied Turing test for view-invariant object recognition
7 Pages. 4 figures, 1 table. This paper was accepted to the CogSci 2023 Conference. (https://cognitivesciencesociety.org/)
null
null
null
cs.AI q-bio.NC
http://creativecommons.org/licenses/by/4.0/
Recent progress in artificial intelligence has renewed interest in building machines that learn like animals. Almost all of the work comparing learning across biological and artificial systems comes from studies where animals and machines received different training data, obscuring whether differences between animals and machines emerged from differences in learning mechanisms versus training data. We present an experimental approach-a "newborn embodied Turing Test"-that allows newborn animals and machines to be raised in the same environments and tested with the same tasks, permitting direct comparison of their learning abilities. To make this platform, we first collected controlled-rearing data from newborn chicks, then performed "digital twin" experiments in which machines were raised in virtual environments that mimicked the rearing conditions of the chicks. We found that (1) machines (deep reinforcement learning agents with intrinsic motivation) can spontaneously develop visually guided preference behavior, akin to imprinting in newborn chicks, and (2) machines are still far from newborn-level performance on object recognition tasks. Almost all of the chicks developed view-invariant object recognition, whereas the machines tended to develop view-dependent recognition. The learning outcomes were also far more constrained in the chicks versus machines. Ultimately, we anticipate that this approach will help researchers develop embodied AI systems that learn like newborn animals.
[ { "created": "Thu, 8 Jun 2023 22:46:31 GMT", "version": "v1" } ]
2023-06-12
[ [ "Pak", "Denizhan", "" ], [ "Lee", "Donsuk", "" ], [ "Wood", "Samantha M. W.", "" ], [ "Wood", "Justin N.", "" ] ]
Recent progress in artificial intelligence has renewed interest in building machines that learn like animals. Almost all of the work comparing learning across biological and artificial systems comes from studies where animals and machines received different training data, obscuring whether differences between animals and machines emerged from differences in learning mechanisms versus training data. We present an experimental approach-a "newborn embodied Turing Test"-that allows newborn animals and machines to be raised in the same environments and tested with the same tasks, permitting direct comparison of their learning abilities. To make this platform, we first collected controlled-rearing data from newborn chicks, then performed "digital twin" experiments in which machines were raised in virtual environments that mimicked the rearing conditions of the chicks. We found that (1) machines (deep reinforcement learning agents with intrinsic motivation) can spontaneously develop visually guided preference behavior, akin to imprinting in newborn chicks, and (2) machines are still far from newborn-level performance on object recognition tasks. Almost all of the chicks developed view-invariant object recognition, whereas the machines tended to develop view-dependent recognition. The learning outcomes were also far more constrained in the chicks versus machines. Ultimately, we anticipate that this approach will help researchers develop embodied AI systems that learn like newborn animals.