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
2003.05266
Adrian Brandemuehl
Leiv Andresen, Adrian Brandemuehl, Alex H\"onger, Benson Kuan, Niclas V\"odisch, Hermann Blum, Victor Reijgwart, Lukas Bernreiter, Lukas Schaupp, Jen Jen Chung, Mathias B\"urki, Martin R. Oswald, Roland Siegwart and Abel Gawel
Accurate Mapping and Planning for Autonomous Racing
null
2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020, pp. 4743-4749
10.1109/IROS45743.2020.9341702
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper presents the perception, mapping, and planning pipeline implemented on an autonomous race car. It was developed by the 2019 AMZ driverless team for the Formula Student Germany (FSG) 2019 driverless competition, where it won 1st place overall. The presented solution combines early fusion of camera and LiDAR data, a layered mapping approach, and a planning approach that uses Bayesian filtering to achieve high-speed driving on unknown race tracks while creating accurate maps. We benchmark the method against our team's previous solution, which won FSG 2018, and show improved accuracy when driving at the same speeds. Furthermore, the new pipeline makes it possible to reliably raise the maximum driving speed in unknown environments from 3~m/s to 12~m/s while still mapping with an acceptable RMSE of 0.29~m.
[ { "created": "Wed, 11 Mar 2020 13:08:21 GMT", "version": "v1" }, { "created": "Thu, 12 Mar 2020 13:32:45 GMT", "version": "v2" }, { "created": "Sat, 1 Aug 2020 15:34:29 GMT", "version": "v3" }, { "created": "Thu, 17 Sep 2020 20:05:34 GMT", "version": "v4" } ]
2021-03-02
[ [ "Andresen", "Leiv", "" ], [ "Brandemuehl", "Adrian", "" ], [ "Hönger", "Alex", "" ], [ "Kuan", "Benson", "" ], [ "Vödisch", "Niclas", "" ], [ "Blum", "Hermann", "" ], [ "Reijgwart", "Victor", "" ], [ "Bernreiter", "Lukas", "" ], [ "Schaupp", "Lukas", "" ], [ "Chung", "Jen Jen", "" ], [ "Bürki", "Mathias", "" ], [ "Oswald", "Martin R.", "" ], [ "Siegwart", "Roland", "" ], [ "Gawel", "Abel", "" ] ]
This paper presents the perception, mapping, and planning pipeline implemented on an autonomous race car. It was developed by the 2019 AMZ driverless team for the Formula Student Germany (FSG) 2019 driverless competition, where it won 1st place overall. The presented solution combines early fusion of camera and LiDAR data, a layered mapping approach, and a planning approach that uses Bayesian filtering to achieve high-speed driving on unknown race tracks while creating accurate maps. We benchmark the method against our team's previous solution, which won FSG 2018, and show improved accuracy when driving at the same speeds. Furthermore, the new pipeline makes it possible to reliably raise the maximum driving speed in unknown environments from 3~m/s to 12~m/s while still mapping with an acceptable RMSE of 0.29~m.
2301.01913
Tom Marty
Tom Marty, Tristan Fran\c{c}ois, Pierre Tessier, Louis Gauthier, Louis-Martin Rousseau, Quentin Cappart
Learning a Generic Value-Selection Heuristic Inside a Constraint Programming Solver
15 pages
Constraint Programming 29 (2023) 25:1--25:19
10.4230/LIPIcs.CP.2023.25
null
cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Constraint programming is known for being an efficient approach for solving combinatorial problems. Important design choices in a solver are the branching heuristics, which are designed to lead the search to the best solutions in a minimum amount of time. However, developing these heuristics is a time-consuming process that requires problem-specific expertise. This observation has motivated many efforts to use machine learning to automatically learn efficient heuristics without expert intervention. To the best of our knowledge, it is still an open research question. Although several generic variable-selection heuristics are available in the literature, the options for a generic value-selection heuristic are more scarce. In this paper, we propose to tackle this issue by introducing a generic learning procedure that can be used to obtain a value-selection heuristic inside a constraint programming solver. This has been achieved thanks to the combination of a deep Q-learning algorithm, a tailored reward signal, and a heterogeneous graph neural network architecture. Experiments on graph coloring, maximum independent set, and maximum cut problems show that our framework is able to find better solutions close to optimality without requiring a large amounts of backtracks while being generic.
[ { "created": "Thu, 5 Jan 2023 05:13:48 GMT", "version": "v1" }, { "created": "Wed, 26 Jul 2023 22:15:08 GMT", "version": "v2" }, { "created": "Mon, 2 Oct 2023 16:59:40 GMT", "version": "v3" } ]
2024-04-17
[ [ "Marty", "Tom", "" ], [ "François", "Tristan", "" ], [ "Tessier", "Pierre", "" ], [ "Gauthier", "Louis", "" ], [ "Rousseau", "Louis-Martin", "" ], [ "Cappart", "Quentin", "" ] ]
Constraint programming is known for being an efficient approach for solving combinatorial problems. Important design choices in a solver are the branching heuristics, which are designed to lead the search to the best solutions in a minimum amount of time. However, developing these heuristics is a time-consuming process that requires problem-specific expertise. This observation has motivated many efforts to use machine learning to automatically learn efficient heuristics without expert intervention. To the best of our knowledge, it is still an open research question. Although several generic variable-selection heuristics are available in the literature, the options for a generic value-selection heuristic are more scarce. In this paper, we propose to tackle this issue by introducing a generic learning procedure that can be used to obtain a value-selection heuristic inside a constraint programming solver. This has been achieved thanks to the combination of a deep Q-learning algorithm, a tailored reward signal, and a heterogeneous graph neural network architecture. Experiments on graph coloring, maximum independent set, and maximum cut problems show that our framework is able to find better solutions close to optimality without requiring a large amounts of backtracks while being generic.
2212.04972
Jialiang Lin
Jialiang Lin, Jiaxin Song, Zhangping Zhou, Yidong Chen, Xiaodong Shi
MOPRD: A multidisciplinary open peer review dataset
Please cite the version of Neural Computing and Applications
Neural Computing and Applications, Vol. 35, Issue 34, pp. 24191-24206 (2023)
10.1007/s00521-023-08891-5
null
cs.DL cs.AI cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Open peer review is a growing trend in academic publications. Public access to peer review data can benefit both the academic and publishing communities. It also serves as a great support to studies on review comment generation and further to the realization of automated scholarly paper review. However, most of the existing peer review datasets do not provide data that cover the whole peer review process. Apart from this, their data are not diversified enough as the data are mainly collected from the field of computer science. These two drawbacks of the currently available peer review datasets need to be addressed to unlock more opportunities for related studies. In response, we construct MOPRD, a multidisciplinary open peer review dataset. This dataset consists of paper metadata, multiple version manuscripts, review comments, meta-reviews, author's rebuttal letters, and editorial decisions. Moreover, we propose a modular guided review comment generation method based on MOPRD. Experiments show that our method delivers better performance as indicated by both automatic metrics and human evaluation. We also explore other potential applications of MOPRD, including meta-review generation, editorial decision prediction, author rebuttal generation, and scientometric analysis. MOPRD is a strong endorsement for further studies in peer review-related research and other applications.
[ { "created": "Fri, 9 Dec 2022 16:35:14 GMT", "version": "v1" }, { "created": "Tue, 14 Nov 2023 18:06:48 GMT", "version": "v2" } ]
2023-11-15
[ [ "Lin", "Jialiang", "" ], [ "Song", "Jiaxin", "" ], [ "Zhou", "Zhangping", "" ], [ "Chen", "Yidong", "" ], [ "Shi", "Xiaodong", "" ] ]
Open peer review is a growing trend in academic publications. Public access to peer review data can benefit both the academic and publishing communities. It also serves as a great support to studies on review comment generation and further to the realization of automated scholarly paper review. However, most of the existing peer review datasets do not provide data that cover the whole peer review process. Apart from this, their data are not diversified enough as the data are mainly collected from the field of computer science. These two drawbacks of the currently available peer review datasets need to be addressed to unlock more opportunities for related studies. In response, we construct MOPRD, a multidisciplinary open peer review dataset. This dataset consists of paper metadata, multiple version manuscripts, review comments, meta-reviews, author's rebuttal letters, and editorial decisions. Moreover, we propose a modular guided review comment generation method based on MOPRD. Experiments show that our method delivers better performance as indicated by both automatic metrics and human evaluation. We also explore other potential applications of MOPRD, including meta-review generation, editorial decision prediction, author rebuttal generation, and scientometric analysis. MOPRD is a strong endorsement for further studies in peer review-related research and other applications.
2306.13381
Rahul Nair
Rahul Nair
Co-creating a globally interpretable model with human input
Paper at Artificial Intelligence & Human-Computer Interaction Workshop at ICML 2023
null
null
null
cs.HC cs.LG
http://creativecommons.org/licenses/by/4.0/
We consider an aggregated human-AI collaboration aimed at generating a joint interpretable model. The model takes the form of Boolean decision rules, where human input is provided in the form of logical conditions or as partial templates. This focus on the combined construction of a model offers a different perspective on joint decision making. Previous efforts have typically focused on aggregating outcomes rather than decisions logic. We demonstrate the proposed approach through two examples and highlight the usefulness and challenges of the approach.
[ { "created": "Fri, 23 Jun 2023 09:03:16 GMT", "version": "v1" } ]
2023-06-26
[ [ "Nair", "Rahul", "" ] ]
We consider an aggregated human-AI collaboration aimed at generating a joint interpretable model. The model takes the form of Boolean decision rules, where human input is provided in the form of logical conditions or as partial templates. This focus on the combined construction of a model offers a different perspective on joint decision making. Previous efforts have typically focused on aggregating outcomes rather than decisions logic. We demonstrate the proposed approach through two examples and highlight the usefulness and challenges of the approach.
1204.4467
I-Hong Hou
I-Hong Hou and Rahul Singh
Real-Time Stochastic Processing Networks with Concurrent Resource Requirements
null
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Stochastic Processing Networks (SPNs) can be used to model communication networks, manufacturing systems, service systems, etc. We consider a real-time SPN where tasks generate jobs with strict deadlines according to their traffic patterns. Each job requires the concurrent usage of some resources to be processed. The processing time of a job may be stochastic, and may not be known until the job completes. Finally, each task may require that some portion of its tasks to be completed on time. In this paper, we study the problem of verifying whether it is feasible to fulfill the requirements of tasks, and of designing scheduling policies that actually fulfill the requirements. We first address these problems for systems where there is only one resource. Such systems are analog to ones studied in a previous work, and, similar to the previous work, we can develop sharp conditions for feasibility and scheduling policy that is feasibility-optimal. We then study systems with two resources where there are jobs that require both resources to be processed. We show that there is a reduction method that turns systems with two resources into equivalent single-resource systems. Based on this method, we can also derive sharp feasibility conditions and feasibility-optimal scheduling policies for systems with two resources.
[ { "created": "Thu, 19 Apr 2012 20:34:57 GMT", "version": "v1" } ]
2012-04-23
[ [ "Hou", "I-Hong", "" ], [ "Singh", "Rahul", "" ] ]
Stochastic Processing Networks (SPNs) can be used to model communication networks, manufacturing systems, service systems, etc. We consider a real-time SPN where tasks generate jobs with strict deadlines according to their traffic patterns. Each job requires the concurrent usage of some resources to be processed. The processing time of a job may be stochastic, and may not be known until the job completes. Finally, each task may require that some portion of its tasks to be completed on time. In this paper, we study the problem of verifying whether it is feasible to fulfill the requirements of tasks, and of designing scheduling policies that actually fulfill the requirements. We first address these problems for systems where there is only one resource. Such systems are analog to ones studied in a previous work, and, similar to the previous work, we can develop sharp conditions for feasibility and scheduling policy that is feasibility-optimal. We then study systems with two resources where there are jobs that require both resources to be processed. We show that there is a reduction method that turns systems with two resources into equivalent single-resource systems. Based on this method, we can also derive sharp feasibility conditions and feasibility-optimal scheduling policies for systems with two resources.
1908.04133
Matthias Niedermaier
Matthias Niedermaier and Dominik Merli and Georg Sigl
A Secure Dual-MCU Architecture for Robust Communication of IIoT Devices
null
2019 8th Mediterranean Conference on Embedded Computing (MECO)
10.1109/MECO.2019.8760188
null
cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Industrial Internet of Things (IIoT) has already become a part of our everyday life be it water supply, smart grid, or production, IIoT is everywhere. For example, factory operators want to know the current state of the production line. These new demands for data acquisition in modern plants require industrial components to be able to communicate. Nowadays, network communication in Industrial Control Systems (ICSs) is often implemented via an IP-based protocol. This intercommunication also brings a larger attack surface for hackers. If an IIoT device is influenced by attackers, the physical process could be affected. For example, a high network load could cause a high Central Processing Unit (CPU) load and influence the reaction time on the physical control side. In this paper, we introduce a dual Microcontroller Unit (MCU) setup to ensure a resilient controlling for IIoT devices like Programmable Logic Controllers (PLCs). We introduce a possible solution for the demand of secure architectures in the IIoT. Moreover, we provide a Proof of Concept (PoC) implementation with a benchmark and a comparison with a standard PLC.
[ { "created": "Mon, 12 Aug 2019 12:58:53 GMT", "version": "v1" } ]
2019-08-13
[ [ "Niedermaier", "Matthias", "" ], [ "Merli", "Dominik", "" ], [ "Sigl", "Georg", "" ] ]
The Industrial Internet of Things (IIoT) has already become a part of our everyday life be it water supply, smart grid, or production, IIoT is everywhere. For example, factory operators want to know the current state of the production line. These new demands for data acquisition in modern plants require industrial components to be able to communicate. Nowadays, network communication in Industrial Control Systems (ICSs) is often implemented via an IP-based protocol. This intercommunication also brings a larger attack surface for hackers. If an IIoT device is influenced by attackers, the physical process could be affected. For example, a high network load could cause a high Central Processing Unit (CPU) load and influence the reaction time on the physical control side. In this paper, we introduce a dual Microcontroller Unit (MCU) setup to ensure a resilient controlling for IIoT devices like Programmable Logic Controllers (PLCs). We introduce a possible solution for the demand of secure architectures in the IIoT. Moreover, we provide a Proof of Concept (PoC) implementation with a benchmark and a comparison with a standard PLC.
2002.12324
Eric Brachmann
Eric Brachmann and Carsten Rother
Visual Camera Re-Localization from RGB and RGB-D Images Using DSAC
null
null
null
null
cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We describe a learning-based system that estimates the camera position and orientation from a single input image relative to a known environment. The system is flexible w.r.t. the amount of information available at test and at training time, catering to different applications. Input images can be RGB-D or RGB, and a 3D model of the environment can be utilized for training but is not necessary. In the minimal case, our system requires only RGB images and ground truth poses at training time, and it requires only a single RGB image at test time. The framework consists of a deep neural network and fully differentiable pose optimization. The neural network predicts so called scene coordinates, i.e. dense correspondences between the input image and 3D scene space of the environment. The pose optimization implements robust fitting of pose parameters using differentiable RANSAC (DSAC) to facilitate end-to-end training. The system, an extension of DSAC++ and referred to as DSAC*, achieves state-of-the-art accuracy an various public datasets for RGB-based re-localization, and competitive accuracy for RGB-D-based re-localization.
[ { "created": "Thu, 27 Feb 2020 18:45:21 GMT", "version": "v1" }, { "created": "Fri, 28 Aug 2020 12:07:47 GMT", "version": "v2" }, { "created": "Mon, 31 Aug 2020 12:29:26 GMT", "version": "v3" }, { "created": "Fri, 9 Oct 2020 15:03:02 GMT", "version": "v4" } ]
2020-10-12
[ [ "Brachmann", "Eric", "" ], [ "Rother", "Carsten", "" ] ]
We describe a learning-based system that estimates the camera position and orientation from a single input image relative to a known environment. The system is flexible w.r.t. the amount of information available at test and at training time, catering to different applications. Input images can be RGB-D or RGB, and a 3D model of the environment can be utilized for training but is not necessary. In the minimal case, our system requires only RGB images and ground truth poses at training time, and it requires only a single RGB image at test time. The framework consists of a deep neural network and fully differentiable pose optimization. The neural network predicts so called scene coordinates, i.e. dense correspondences between the input image and 3D scene space of the environment. The pose optimization implements robust fitting of pose parameters using differentiable RANSAC (DSAC) to facilitate end-to-end training. The system, an extension of DSAC++ and referred to as DSAC*, achieves state-of-the-art accuracy an various public datasets for RGB-based re-localization, and competitive accuracy for RGB-D-based re-localization.
2310.17417
John Liu
Hing Lie, Kachina Studer, Zhen Zhao, Ben Thomson, Dishita G Turakhia, John Liu
Training for Open-Ended Drilling through a Virtual Reality Simulation
10 pages, 4 figures, 9 tables
null
null
null
cs.HC
http://creativecommons.org/licenses/by/4.0/
Virtual Reality (VR) can support effective and scalable training of psychomotor skills in manufacturing. However, many industry training modules offer experiences that are close-ended and do not allow for human error. We aim to address this gap in VR training tools for psychomotor skills training by exploring an open-ended approach to the system design. We designed a VR training simulation prototype to perform open-ended practice of drilling using a 3-axis milling machine. The simulation employs near "end-to-end" instruction through a safety module, a setup and drilling tutorial, open-ended practice complete with warnings of mistakes and failures, and a function to assess the geometries and locations of drilled holes against an engineering drawing. We developed and conducted a user study within an undergraduate-level introductory fabrication course to investigate the impact of open-ended VR practice on learning outcomes. Study results reveal positive trends, with the VR group successfully completing the machining task of drilling at a higher rate (75% vs 64%), with fewer mistakes (1.75 vs 2.14 score), and in less time (17.67 mins vs 21.57 mins) compared to the control group. We discuss our findings and limitations and implications for the design of open-ended VR training systems for learning psychomotor skills.
[ { "created": "Thu, 26 Oct 2023 14:22:30 GMT", "version": "v1" } ]
2023-10-27
[ [ "Lie", "Hing", "" ], [ "Studer", "Kachina", "" ], [ "Zhao", "Zhen", "" ], [ "Thomson", "Ben", "" ], [ "Turakhia", "Dishita G", "" ], [ "Liu", "John", "" ] ]
Virtual Reality (VR) can support effective and scalable training of psychomotor skills in manufacturing. However, many industry training modules offer experiences that are close-ended and do not allow for human error. We aim to address this gap in VR training tools for psychomotor skills training by exploring an open-ended approach to the system design. We designed a VR training simulation prototype to perform open-ended practice of drilling using a 3-axis milling machine. The simulation employs near "end-to-end" instruction through a safety module, a setup and drilling tutorial, open-ended practice complete with warnings of mistakes and failures, and a function to assess the geometries and locations of drilled holes against an engineering drawing. We developed and conducted a user study within an undergraduate-level introductory fabrication course to investigate the impact of open-ended VR practice on learning outcomes. Study results reveal positive trends, with the VR group successfully completing the machining task of drilling at a higher rate (75% vs 64%), with fewer mistakes (1.75 vs 2.14 score), and in less time (17.67 mins vs 21.57 mins) compared to the control group. We discuss our findings and limitations and implications for the design of open-ended VR training systems for learning psychomotor skills.
2406.00588
Xiao-Shan Gao
Lijia Yu and Shuang Liu and Yibo Miao and Xiao-Shan Gao and Lijun Zhang
Generalization Bound and New Algorithm for Clean-Label Backdoor Attack
null
null
null
null
cs.LG cs.CR math.ST stat.TH
http://creativecommons.org/licenses/by/4.0/
The generalization bound is a crucial theoretical tool for assessing the generalizability of learning methods and there exist vast literatures on generalizability of normal learning, adversarial learning, and data poisoning. Unlike other data poison attacks, the backdoor attack has the special property that the poisoned triggers are contained in both the training set and the test set and the purpose of the attack is two-fold. To our knowledge, the generalization bound for the backdoor attack has not been established. In this paper, we fill this gap by deriving algorithm-independent generalization bounds in the clean-label backdoor attack scenario. Precisely, based on the goals of backdoor attack, we give upper bounds for the clean sample population errors and the poison population errors in terms of the empirical error on the poisoned training dataset. Furthermore, based on the theoretical result, a new clean-label backdoor attack is proposed that computes the poisoning trigger by combining adversarial noise and indiscriminate poison. We show its effectiveness in a variety of settings.
[ { "created": "Sun, 2 Jun 2024 01:46:58 GMT", "version": "v1" } ]
2024-06-05
[ [ "Yu", "Lijia", "" ], [ "Liu", "Shuang", "" ], [ "Miao", "Yibo", "" ], [ "Gao", "Xiao-Shan", "" ], [ "Zhang", "Lijun", "" ] ]
The generalization bound is a crucial theoretical tool for assessing the generalizability of learning methods and there exist vast literatures on generalizability of normal learning, adversarial learning, and data poisoning. Unlike other data poison attacks, the backdoor attack has the special property that the poisoned triggers are contained in both the training set and the test set and the purpose of the attack is two-fold. To our knowledge, the generalization bound for the backdoor attack has not been established. In this paper, we fill this gap by deriving algorithm-independent generalization bounds in the clean-label backdoor attack scenario. Precisely, based on the goals of backdoor attack, we give upper bounds for the clean sample population errors and the poison population errors in terms of the empirical error on the poisoned training dataset. Furthermore, based on the theoretical result, a new clean-label backdoor attack is proposed that computes the poisoning trigger by combining adversarial noise and indiscriminate poison. We show its effectiveness in a variety of settings.
1511.02919
Deepti Ghadiyaram
Deepti Ghadiyaram and Alan C. Bovik
Massive Online Crowdsourced Study of Subjective and Objective Picture Quality
16 pages
null
10.1109/TIP.2015.2500021
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Most publicly available image quality databases have been created under highly controlled conditions by introducing graded simulated distortions onto high-quality photographs. However, images captured using typical real-world mobile camera devices are usually afflicted by complex mixtures of multiple distortions, which are not necessarily well-modeled by the synthetic distortions found in existing databases. The originators of existing legacy databases usually conducted human psychometric studies to obtain statistically meaningful sets of human opinion scores on images in a stringently controlled visual environment, resulting in small data collections relative to other kinds of image analysis databases. Towards overcoming these limitations, we designed and created a new database that we call the LIVE In the Wild Image Quality Challenge Database, which contains widely diverse authentic image distortions on a large number of images captured using a representative variety of modern mobile devices. We also designed and implemented a new online crowdsourcing system, which we have used to conduct a very large-scale, multi-month image quality assessment subjective study. Our database consists of over 350000 opinion scores on 1162 images evaluated by over 7000 unique human observers. Despite the lack of control over the experimental environments of the numerous study participants, we demonstrate excellent internal consistency of the subjective dataset. We also evaluate several top-performing blind Image Quality Assessment algorithms on it and present insights on how mixtures of distortions challenge both end users as well as automatic perceptual quality prediction models.
[ { "created": "Mon, 9 Nov 2015 22:39:58 GMT", "version": "v1" } ]
2016-01-20
[ [ "Ghadiyaram", "Deepti", "" ], [ "Bovik", "Alan C.", "" ] ]
Most publicly available image quality databases have been created under highly controlled conditions by introducing graded simulated distortions onto high-quality photographs. However, images captured using typical real-world mobile camera devices are usually afflicted by complex mixtures of multiple distortions, which are not necessarily well-modeled by the synthetic distortions found in existing databases. The originators of existing legacy databases usually conducted human psychometric studies to obtain statistically meaningful sets of human opinion scores on images in a stringently controlled visual environment, resulting in small data collections relative to other kinds of image analysis databases. Towards overcoming these limitations, we designed and created a new database that we call the LIVE In the Wild Image Quality Challenge Database, which contains widely diverse authentic image distortions on a large number of images captured using a representative variety of modern mobile devices. We also designed and implemented a new online crowdsourcing system, which we have used to conduct a very large-scale, multi-month image quality assessment subjective study. Our database consists of over 350000 opinion scores on 1162 images evaluated by over 7000 unique human observers. Despite the lack of control over the experimental environments of the numerous study participants, we demonstrate excellent internal consistency of the subjective dataset. We also evaluate several top-performing blind Image Quality Assessment algorithms on it and present insights on how mixtures of distortions challenge both end users as well as automatic perceptual quality prediction models.
2206.01970
Haosen Wang
Enqiang Zhu, Haosen Wang, Yu Zhang, Kai Zhang, Chanjuan Liu
PHEE: A phased hybrid evaluation-enhanced approach for identifying influential users in social networks
null
null
null
null
cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
For the purpose of maximizing the spread of influence caused by a certain small number k of nodes in a social network, we are asked to find a k-subset of nodes (i.e., a seed set) with the best capacity to influence the nodes not in it. This problem of influence maximization (IM) has wide application, belongs to subset problems, and is NP-hard. To solve it, we should theoretically examine all seed sets and evaluate their influence spreads, which is time-consuming. Therefore, metaheuristic strategies are generally employed to gain a good seed set within a reasonable time. We observe that many algorithms for the IM problem only adopt a uniform mechanism in the whole solution search process, which lacks a response measure when the algorithm becomes trapped in a local optimum. To address this issue, we propose a phased hybrid evaluation-enhanced (PHEE) approach for IM, which utilizes two distinct search strategies to enhance the search of optimal solutions: a randomized range division evolutionary (RandRDE) algorithm to improve the solution quality, and a fast convergence strategy. Our approach is evaluated on 10 real-world social networks of different sizes and types. Experimental results demonstrate that our algorithm is efficient and obtains the best influence spread for all the datasets compared with three state-of-the-art algorithms, outperforms the time consuming CELF algorithm on four datasets, and performs worse than CELF on only two networks.
[ { "created": "Sat, 4 Jun 2022 12:00:08 GMT", "version": "v1" } ]
2022-06-07
[ [ "Zhu", "Enqiang", "" ], [ "Wang", "Haosen", "" ], [ "Zhang", "Yu", "" ], [ "Zhang", "Kai", "" ], [ "Liu", "Chanjuan", "" ] ]
For the purpose of maximizing the spread of influence caused by a certain small number k of nodes in a social network, we are asked to find a k-subset of nodes (i.e., a seed set) with the best capacity to influence the nodes not in it. This problem of influence maximization (IM) has wide application, belongs to subset problems, and is NP-hard. To solve it, we should theoretically examine all seed sets and evaluate their influence spreads, which is time-consuming. Therefore, metaheuristic strategies are generally employed to gain a good seed set within a reasonable time. We observe that many algorithms for the IM problem only adopt a uniform mechanism in the whole solution search process, which lacks a response measure when the algorithm becomes trapped in a local optimum. To address this issue, we propose a phased hybrid evaluation-enhanced (PHEE) approach for IM, which utilizes two distinct search strategies to enhance the search of optimal solutions: a randomized range division evolutionary (RandRDE) algorithm to improve the solution quality, and a fast convergence strategy. Our approach is evaluated on 10 real-world social networks of different sizes and types. Experimental results demonstrate that our algorithm is efficient and obtains the best influence spread for all the datasets compared with three state-of-the-art algorithms, outperforms the time consuming CELF algorithm on four datasets, and performs worse than CELF on only two networks.
2203.10581
Ariel Gera
Eyal Shnarch, Ariel Gera, Alon Halfon, Lena Dankin, Leshem Choshen, Ranit Aharonov, Noam Slonim
Cluster & Tune: Boost Cold Start Performance in Text Classification
9 pages, 6 figures; To be published in ACL 2022
null
null
null
cs.CL cs.LG
http://creativecommons.org/licenses/by-nc-nd/4.0/
In real-world scenarios, a text classification task often begins with a cold start, when labeled data is scarce. In such cases, the common practice of fine-tuning pre-trained models, such as BERT, for a target classification task, is prone to produce poor performance. We suggest a method to boost the performance of such models by adding an intermediate unsupervised classification task, between the pre-training and fine-tuning phases. As such an intermediate task, we perform clustering and train the pre-trained model on predicting the cluster labels. We test this hypothesis on various data sets, and show that this additional classification phase can significantly improve performance, mainly for topical classification tasks, when the number of labeled instances available for fine-tuning is only a couple of dozen to a few hundred.
[ { "created": "Sun, 20 Mar 2022 15:29:34 GMT", "version": "v1" } ]
2022-03-22
[ [ "Shnarch", "Eyal", "" ], [ "Gera", "Ariel", "" ], [ "Halfon", "Alon", "" ], [ "Dankin", "Lena", "" ], [ "Choshen", "Leshem", "" ], [ "Aharonov", "Ranit", "" ], [ "Slonim", "Noam", "" ] ]
In real-world scenarios, a text classification task often begins with a cold start, when labeled data is scarce. In such cases, the common practice of fine-tuning pre-trained models, such as BERT, for a target classification task, is prone to produce poor performance. We suggest a method to boost the performance of such models by adding an intermediate unsupervised classification task, between the pre-training and fine-tuning phases. As such an intermediate task, we perform clustering and train the pre-trained model on predicting the cluster labels. We test this hypothesis on various data sets, and show that this additional classification phase can significantly improve performance, mainly for topical classification tasks, when the number of labeled instances available for fine-tuning is only a couple of dozen to a few hundred.
2012.01644
Joy Hsu
Joy Hsu, Jeffrey Gu, Gong-Her Wu, Wah Chiu, Serena Yeung
Capturing implicit hierarchical structure in 3D biomedical images with self-supervised hyperbolic representations
To appear at NeurIPS 2021
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the task of representation learning for unsupervised segmentation of 3D voxel-grid biomedical images. We show that models that capture implicit hierarchical relationships between subvolumes are better suited for this task. To that end, we consider encoder-decoder architectures with a hyperbolic latent space, to explicitly capture hierarchical relationships present in subvolumes of the data. We propose utilizing a 3D hyperbolic variational autoencoder with a novel gyroplane convolutional layer to map from the embedding space back to 3D images. To capture these relationships, we introduce an essential self-supervised loss -- in addition to the standard VAE loss -- which infers approximate hierarchies and encourages implicitly related subvolumes to be mapped closer in the embedding space. We present experiments on both synthetic data and biomedical data to validate our hypothesis.
[ { "created": "Thu, 3 Dec 2020 02:15:31 GMT", "version": "v1" }, { "created": "Fri, 4 Dec 2020 23:28:46 GMT", "version": "v2" }, { "created": "Mon, 25 Oct 2021 22:36:34 GMT", "version": "v3" } ]
2021-10-27
[ [ "Hsu", "Joy", "" ], [ "Gu", "Jeffrey", "" ], [ "Wu", "Gong-Her", "" ], [ "Chiu", "Wah", "" ], [ "Yeung", "Serena", "" ] ]
We consider the task of representation learning for unsupervised segmentation of 3D voxel-grid biomedical images. We show that models that capture implicit hierarchical relationships between subvolumes are better suited for this task. To that end, we consider encoder-decoder architectures with a hyperbolic latent space, to explicitly capture hierarchical relationships present in subvolumes of the data. We propose utilizing a 3D hyperbolic variational autoencoder with a novel gyroplane convolutional layer to map from the embedding space back to 3D images. To capture these relationships, we introduce an essential self-supervised loss -- in addition to the standard VAE loss -- which infers approximate hierarchies and encourages implicitly related subvolumes to be mapped closer in the embedding space. We present experiments on both synthetic data and biomedical data to validate our hypothesis.
1807.02816
The-Hien Dang-Ha
The-Hien Dang-Ha
Improving Deep Learning through Automatic Programming
Master's thesis (2014)
null
null
null
cs.LG stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Deep learning and deep architectures are emerging as the best machine learning methods so far in many practical applications such as reducing the dimensionality of data, image classification, speech recognition or object segmentation. In fact, many leading technology companies such as Google, Microsoft or IBM are researching and using deep architectures in their systems to replace other traditional models. Therefore, improving the performance of these models could make a strong impact in the area of machine learning. However, deep learning is a very fast-growing research domain with many core methodologies and paradigms just discovered over the last few years. This thesis will first serve as a short summary of deep learning, which tries to include all of the most important ideas in this research area. Based on this knowledge, we suggested, and conducted some experiments to investigate the possibility of improving the deep learning based on automatic programming (ADATE). Although our experiments did produce good results, there are still many more possibilities that we could not try due to limited time as well as some limitations of the current ADATE version. I hope that this thesis can promote future work on this topic, especially when the next version of ADATE comes out. This thesis also includes a short analysis of the power of ADATE system, which could be useful for other researchers who want to know what it is capable of.
[ { "created": "Sun, 8 Jul 2018 13:38:21 GMT", "version": "v1" } ]
2018-07-10
[ [ "Dang-Ha", "The-Hien", "" ] ]
Deep learning and deep architectures are emerging as the best machine learning methods so far in many practical applications such as reducing the dimensionality of data, image classification, speech recognition or object segmentation. In fact, many leading technology companies such as Google, Microsoft or IBM are researching and using deep architectures in their systems to replace other traditional models. Therefore, improving the performance of these models could make a strong impact in the area of machine learning. However, deep learning is a very fast-growing research domain with many core methodologies and paradigms just discovered over the last few years. This thesis will first serve as a short summary of deep learning, which tries to include all of the most important ideas in this research area. Based on this knowledge, we suggested, and conducted some experiments to investigate the possibility of improving the deep learning based on automatic programming (ADATE). Although our experiments did produce good results, there are still many more possibilities that we could not try due to limited time as well as some limitations of the current ADATE version. I hope that this thesis can promote future work on this topic, especially when the next version of ADATE comes out. This thesis also includes a short analysis of the power of ADATE system, which could be useful for other researchers who want to know what it is capable of.
0911.3306
Pascal Urso
Nacer Boudjlida (INRIA Lorraine - LORIA), Jean-Pierre Jacquot (LORIA), Pascal Urso (INRIA Lorraine - LORIA)
Software Engineering Education by Example
null
5th China - Europe International Symposium on Software Industry Oriented Education (CEISIE 2009), Bordeaux : France (2009)
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Based on the old but famous distinction between "in the small" and "in the large" software development, at Nancy Universit\'e, UHP Nancy 1, we experience for a while software engineering education thanks to actual project engineering. This education method has the merit to enable students to discover and to overcome actual problems when faced to a large project which may be conducted by a large development team. The mode of education is a simulation of an actual software engineering project as encountered in "real life\'e" activities.
[ { "created": "Tue, 17 Nov 2009 13:36:59 GMT", "version": "v1" } ]
2009-11-18
[ [ "Boudjlida", "Nacer", "", "INRIA Lorraine - LORIA" ], [ "Jacquot", "Jean-Pierre", "", "LORIA" ], [ "Urso", "Pascal", "", "INRIA Lorraine - LORIA" ] ]
Based on the old but famous distinction between "in the small" and "in the large" software development, at Nancy Universit\'e, UHP Nancy 1, we experience for a while software engineering education thanks to actual project engineering. This education method has the merit to enable students to discover and to overcome actual problems when faced to a large project which may be conducted by a large development team. The mode of education is a simulation of an actual software engineering project as encountered in "real life\'e" activities.
2012.01172
Burak Yildiz
Burak Yildiz, Hayley Hung, Jesse H. Krijthe, Cynthia C. S. Liem, Marco Loog, Gosia Migut, Frans Oliehoek, Annibale Panichella, Przemyslaw Pawelczak, Stjepan Picek, Mathijs de Weerdt, and Jan van Gemert
ReproducedPapers.org: Openly teaching and structuring machine learning reproducibility
Accepted to RRPR 2020: Third Workshop on Reproducible Research in Pattern Recognition
null
10.1007/978-3-030-76423-4_1
null
cs.CY cs.CV
http://creativecommons.org/licenses/by/4.0/
We present ReproducedPapers.org: an open online repository for teaching and structuring machine learning reproducibility. We evaluate doing a reproduction project among students and the added value of an online reproduction repository among AI researchers. We use anonymous self-assessment surveys and obtained 144 responses. Results suggest that students who do a reproduction project place more value on scientific reproductions and become more critical thinkers. Students and AI researchers agree that our online reproduction repository is valuable.
[ { "created": "Tue, 1 Dec 2020 11:19:45 GMT", "version": "v1" } ]
2021-06-11
[ [ "Yildiz", "Burak", "" ], [ "Hung", "Hayley", "" ], [ "Krijthe", "Jesse H.", "" ], [ "Liem", "Cynthia C. S.", "" ], [ "Loog", "Marco", "" ], [ "Migut", "Gosia", "" ], [ "Oliehoek", "Frans", "" ], [ "Panichella", "Annibale", "" ], [ "Pawelczak", "Przemyslaw", "" ], [ "Picek", "Stjepan", "" ], [ "de Weerdt", "Mathijs", "" ], [ "van Gemert", "Jan", "" ] ]
We present ReproducedPapers.org: an open online repository for teaching and structuring machine learning reproducibility. We evaluate doing a reproduction project among students and the added value of an online reproduction repository among AI researchers. We use anonymous self-assessment surveys and obtained 144 responses. Results suggest that students who do a reproduction project place more value on scientific reproductions and become more critical thinkers. Students and AI researchers agree that our online reproduction repository is valuable.
1203.4627
Vasilis Gkatzelis
Richard Cole, Vasilis Gkatzelis, Gagan Goel
Truthfulness, Proportional Fairness, and Efficiency
null
null
null
null
cs.GT cs.DS cs.MA
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
How does one allocate a collection of resources to a set of strategic agents in a fair and efficient manner without using money? For in many scenarios it is not feasible to use money to compensate agents for otherwise unsatisfactory outcomes. This paper studies this question, looking at both fairness and efficiency measures. We employ the proportionally fair solution, which is a well-known fairness concept for money-free settings. But although finding a proportionally fair solution is computationally tractable, it cannot be implemented in a truthful fashion. Consequently, we seek approximate solutions. We give several truthful mechanisms which achieve proportional fairness in an approximate sense. We use a strong notion of approximation, requiring the mechanism to give each agent a good approximation of its proportionally fair utility. In particular, one of our mechanisms provides a better and better approximation factor as the minimum demand for every good increases. A motivating example is provided by the massive privatization auction in the Czech republic in the early 90s. With regard to efficiency, prior work has shown a lower bound of 0.5 on the approximation factor of any swap-dictatorial mechanism approximating a social welfare measure even for the two agents and multiple goods case. We surpass this lower bound by designing a non-swap-dictatorial mechanism for this case. Interestingly, the new mechanism builds on the notion of proportional fairness.
[ { "created": "Tue, 20 Mar 2012 23:55:49 GMT", "version": "v1" }, { "created": "Fri, 6 Jul 2012 22:16:17 GMT", "version": "v2" } ]
2012-07-10
[ [ "Cole", "Richard", "" ], [ "Gkatzelis", "Vasilis", "" ], [ "Goel", "Gagan", "" ] ]
How does one allocate a collection of resources to a set of strategic agents in a fair and efficient manner without using money? For in many scenarios it is not feasible to use money to compensate agents for otherwise unsatisfactory outcomes. This paper studies this question, looking at both fairness and efficiency measures. We employ the proportionally fair solution, which is a well-known fairness concept for money-free settings. But although finding a proportionally fair solution is computationally tractable, it cannot be implemented in a truthful fashion. Consequently, we seek approximate solutions. We give several truthful mechanisms which achieve proportional fairness in an approximate sense. We use a strong notion of approximation, requiring the mechanism to give each agent a good approximation of its proportionally fair utility. In particular, one of our mechanisms provides a better and better approximation factor as the minimum demand for every good increases. A motivating example is provided by the massive privatization auction in the Czech republic in the early 90s. With regard to efficiency, prior work has shown a lower bound of 0.5 on the approximation factor of any swap-dictatorial mechanism approximating a social welfare measure even for the two agents and multiple goods case. We surpass this lower bound by designing a non-swap-dictatorial mechanism for this case. Interestingly, the new mechanism builds on the notion of proportional fairness.
2404.00987
Ruowen Zhao
Ruowen Zhao, Zhengyi Wang, Yikai Wang, Zihan Zhou and Jun Zhu
FlexiDreamer: Single Image-to-3D Generation with FlexiCubes
Project page: https://flexidreamer.github.io
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
3D content generation has wide applications in various fields. One of its dominant paradigms is by sparse-view reconstruction using multi-view images generated by diffusion models. However, since directly reconstructing triangle meshes from multi-view images is challenging, most methodologies opt to an implicit representation (such as NeRF) during the sparse-view reconstruction and acquire the target mesh by a post-processing extraction. However, the implicit representation takes extensive time to train and the post-extraction also leads to undesirable visual artifacts. In this paper, we propose FlexiDreamer, a novel framework that directly reconstructs high-quality meshes from multi-view generated images. We utilize an advanced gradient-based mesh optimization, namely FlexiCubes, for multi-view mesh reconstruction, which enables us to generate 3D meshes in an end-to-end manner. To address the reconstruction artifacts owing to the inconsistencies from generated images, we design a hybrid positional encoding scheme to improve the reconstruction geometry and an orientation-aware texture mapping to mitigate surface ghosting. To further enhance the results, we respectively incorporate eikonal and smooth regularizations to reduce geometric holes and surface noise. Our approach can generate high-fidelity 3D meshes in the single image-to-3D downstream task with approximately 1 minute, significantly outperforming previous methods.
[ { "created": "Mon, 1 Apr 2024 08:20:18 GMT", "version": "v1" }, { "created": "Mon, 27 May 2024 09:51:37 GMT", "version": "v2" } ]
2024-05-28
[ [ "Zhao", "Ruowen", "" ], [ "Wang", "Zhengyi", "" ], [ "Wang", "Yikai", "" ], [ "Zhou", "Zihan", "" ], [ "Zhu", "Jun", "" ] ]
3D content generation has wide applications in various fields. One of its dominant paradigms is by sparse-view reconstruction using multi-view images generated by diffusion models. However, since directly reconstructing triangle meshes from multi-view images is challenging, most methodologies opt to an implicit representation (such as NeRF) during the sparse-view reconstruction and acquire the target mesh by a post-processing extraction. However, the implicit representation takes extensive time to train and the post-extraction also leads to undesirable visual artifacts. In this paper, we propose FlexiDreamer, a novel framework that directly reconstructs high-quality meshes from multi-view generated images. We utilize an advanced gradient-based mesh optimization, namely FlexiCubes, for multi-view mesh reconstruction, which enables us to generate 3D meshes in an end-to-end manner. To address the reconstruction artifacts owing to the inconsistencies from generated images, we design a hybrid positional encoding scheme to improve the reconstruction geometry and an orientation-aware texture mapping to mitigate surface ghosting. To further enhance the results, we respectively incorporate eikonal and smooth regularizations to reduce geometric holes and surface noise. Our approach can generate high-fidelity 3D meshes in the single image-to-3D downstream task with approximately 1 minute, significantly outperforming previous methods.
1712.09509
Zhiqing Sun
Zhiqing Sun, Gehui Shen, Zhihong Deng
A Gap-Based Framework for Chinese Word Segmentation via Very Deep Convolutional Networks
Under review as a conference paper at ACL 2018; 10 pages
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Most previous approaches to Chinese word segmentation can be roughly classified into character-based and word-based methods. The former regards this task as a sequence-labeling problem, while the latter directly segments character sequence into words. However, if we consider segmenting a given sentence, the most intuitive idea is to predict whether to segment for each gap between two consecutive characters, which in comparison makes previous approaches seem too complex. Therefore, in this paper, we propose a gap-based framework to implement this intuitive idea. Moreover, very deep convolutional neural networks, namely, ResNets and DenseNets, are exploited in our experiments. Results show that our approach outperforms the best character-based and word-based methods on 5 benchmarks, without any further post-processing module (e.g. Conditional Random Fields) nor beam search.
[ { "created": "Wed, 27 Dec 2017 06:44:02 GMT", "version": "v1" } ]
2017-12-29
[ [ "Sun", "Zhiqing", "" ], [ "Shen", "Gehui", "" ], [ "Deng", "Zhihong", "" ] ]
Most previous approaches to Chinese word segmentation can be roughly classified into character-based and word-based methods. The former regards this task as a sequence-labeling problem, while the latter directly segments character sequence into words. However, if we consider segmenting a given sentence, the most intuitive idea is to predict whether to segment for each gap between two consecutive characters, which in comparison makes previous approaches seem too complex. Therefore, in this paper, we propose a gap-based framework to implement this intuitive idea. Moreover, very deep convolutional neural networks, namely, ResNets and DenseNets, are exploited in our experiments. Results show that our approach outperforms the best character-based and word-based methods on 5 benchmarks, without any further post-processing module (e.g. Conditional Random Fields) nor beam search.
1710.08986
Dimitri Scheftelowitsch
Dimitri Scheftelowitsch, Peter Buchholz, Vahid Hashemi, Holger Hermanns
Multi-Objective Approaches to Markov Decision Processes with Uncertain Transition Parameters
9 pages, 5 figures, preprint for VALUETOOLS 2017
null
null
null
cs.AI cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Markov decision processes (MDPs) are a popular model for performance analysis and optimization of stochastic systems. The parameters of stochastic behavior of MDPs are estimates from empirical observations of a system; their values are not known precisely. Different types of MDPs with uncertain, imprecise or bounded transition rates or probabilities and rewards exist in the literature. Commonly, analysis of models with uncertainties amounts to searching for the most robust policy which means that the goal is to generate a policy with the greatest lower bound on performance (or, symmetrically, the lowest upper bound on costs). However, hedging against an unlikely worst case may lead to losses in other situations. In general, one is interested in policies that behave well in all situations which results in a multi-objective view on decision making. In this paper, we consider policies for the expected discounted reward measure of MDPs with uncertain parameters. In particular, the approach is defined for bounded-parameter MDPs (BMDPs) [8]. In this setting the worst, best and average case performances of a policy are analyzed simultaneously, which yields a multi-scenario multi-objective optimization problem. The paper presents and evaluates approaches to compute the pure Pareto optimal policies in the value vector space.
[ { "created": "Fri, 20 Oct 2017 07:47:41 GMT", "version": "v1" } ]
2017-10-26
[ [ "Scheftelowitsch", "Dimitri", "" ], [ "Buchholz", "Peter", "" ], [ "Hashemi", "Vahid", "" ], [ "Hermanns", "Holger", "" ] ]
Markov decision processes (MDPs) are a popular model for performance analysis and optimization of stochastic systems. The parameters of stochastic behavior of MDPs are estimates from empirical observations of a system; their values are not known precisely. Different types of MDPs with uncertain, imprecise or bounded transition rates or probabilities and rewards exist in the literature. Commonly, analysis of models with uncertainties amounts to searching for the most robust policy which means that the goal is to generate a policy with the greatest lower bound on performance (or, symmetrically, the lowest upper bound on costs). However, hedging against an unlikely worst case may lead to losses in other situations. In general, one is interested in policies that behave well in all situations which results in a multi-objective view on decision making. In this paper, we consider policies for the expected discounted reward measure of MDPs with uncertain parameters. In particular, the approach is defined for bounded-parameter MDPs (BMDPs) [8]. In this setting the worst, best and average case performances of a policy are analyzed simultaneously, which yields a multi-scenario multi-objective optimization problem. The paper presents and evaluates approaches to compute the pure Pareto optimal policies in the value vector space.
1612.05878
Suzhi Bi
Suzhi Bi and Ying Jun Zhang
Graph-based Cyber Security Analysis of State Estimation in Smart Power Grid
This article has been accepted for publication by IEEE Communications Magazine (Dec 2016)
null
null
null
cs.SY cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Smart power grid enables intelligent automation at all levels of power system operation, from electricity generation at power plants to power usage at households. The key enabling factor of an efficient smart grid is its built-in information and communication technology (ICT) that monitors the real-time system operating state and makes control decisions accordingly. As an important building block of the ICT system, power system state estimation is of critical importance to maintain normal operation of the smart grid, which, however, is under mounting threat from potential cyber attacks. In this article, we introduce a graph-based framework for performing cyber-security analysis in power system state estimation. Compared to conventional arithmetic-based security analysis, the graphical characterization of state estimation security provides intuitive visualization of some complex problem structures and enables efficient graphical solution algorithms, which are useful for both defending and attacking the ICT system of smart grid. We also highlight several promising future research directions on graph-based security analysis and its applications in smart power grid.
[ { "created": "Sun, 18 Dec 2016 09:27:40 GMT", "version": "v1" } ]
2016-12-20
[ [ "Bi", "Suzhi", "" ], [ "Zhang", "Ying Jun", "" ] ]
Smart power grid enables intelligent automation at all levels of power system operation, from electricity generation at power plants to power usage at households. The key enabling factor of an efficient smart grid is its built-in information and communication technology (ICT) that monitors the real-time system operating state and makes control decisions accordingly. As an important building block of the ICT system, power system state estimation is of critical importance to maintain normal operation of the smart grid, which, however, is under mounting threat from potential cyber attacks. In this article, we introduce a graph-based framework for performing cyber-security analysis in power system state estimation. Compared to conventional arithmetic-based security analysis, the graphical characterization of state estimation security provides intuitive visualization of some complex problem structures and enables efficient graphical solution algorithms, which are useful for both defending and attacking the ICT system of smart grid. We also highlight several promising future research directions on graph-based security analysis and its applications in smart power grid.
2203.16875
Xiangjun Gao
Xiangjun Gao, Jiaolong Yang, Jongyoo Kim, Sida Peng, Zicheng Liu, Xin Tong
MPS-NeRF: Generalizable 3D Human Rendering from Multiview Images
null
null
null
null
cs.CV cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
There has been rapid progress recently on 3D human rendering, including novel view synthesis and pose animation, based on the advances of neural radiance fields (NeRF). However, most existing methods focus on person-specific training and their training typically requires multi-view videos. This paper deals with a new challenging task -- rendering novel views and novel poses for a person unseen in training, using only multiview images as input. For this task, we propose a simple yet effective method to train a generalizable NeRF with multiview images as conditional input. The key ingredient is a dedicated representation combining a canonical NeRF and a volume deformation scheme. Using a canonical space enables our method to learn shared properties of human and easily generalize to different people. Volume deformation is used to connect the canonical space with input and target images and query image features for radiance and density prediction. We leverage the parametric 3D human model fitted on the input images to derive the deformation, which works quite well in practice when combined with our canonical NeRF. The experiments on both real and synthetic data with the novel view synthesis and pose animation tasks collectively demonstrate the efficacy of our method.
[ { "created": "Thu, 31 Mar 2022 08:09:03 GMT", "version": "v1" }, { "created": "Wed, 27 Jul 2022 06:10:50 GMT", "version": "v2" } ]
2022-07-28
[ [ "Gao", "Xiangjun", "" ], [ "Yang", "Jiaolong", "" ], [ "Kim", "Jongyoo", "" ], [ "Peng", "Sida", "" ], [ "Liu", "Zicheng", "" ], [ "Tong", "Xin", "" ] ]
There has been rapid progress recently on 3D human rendering, including novel view synthesis and pose animation, based on the advances of neural radiance fields (NeRF). However, most existing methods focus on person-specific training and their training typically requires multi-view videos. This paper deals with a new challenging task -- rendering novel views and novel poses for a person unseen in training, using only multiview images as input. For this task, we propose a simple yet effective method to train a generalizable NeRF with multiview images as conditional input. The key ingredient is a dedicated representation combining a canonical NeRF and a volume deformation scheme. Using a canonical space enables our method to learn shared properties of human and easily generalize to different people. Volume deformation is used to connect the canonical space with input and target images and query image features for radiance and density prediction. We leverage the parametric 3D human model fitted on the input images to derive the deformation, which works quite well in practice when combined with our canonical NeRF. The experiments on both real and synthetic data with the novel view synthesis and pose animation tasks collectively demonstrate the efficacy of our method.
2311.10388
Junjie Zhao
Junjie Zhao and Xiang Chen and Guang Yang and Yiheng Shen
Automatic Smart Contract Comment Generation via Large Language Models and In-Context Learning
null
null
null
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The previous smart contract code comment (SCC) generation approaches can be divided into two categories: fine-tuning paradigm-based approaches and information retrieval-based approaches. However, for the fine-tuning paradigm-based approaches, the performance may be limited by the quality of the gathered dataset for the downstream task and they may have knowledge-forgetting issues. While for the information retrieval-based approaches, it is difficult for them to generate high-quality comments if similar code does not exist in the historical repository. Therefore we want to utilize the domain knowledge related to SCC generation in large language models (LLMs) to alleviate the disadvantages of these two types of approaches. In this study, we propose an approach SCCLLM based on LLMs and in-context learning. Specifically, in the demonstration selection phase, SCCLLM retrieves the top-k code snippets from the historical corpus by considering syntax, semantics, and lexical information. In the in-context learning phase, SCCLLM utilizes the retrieved code snippets as demonstrations, which can help to utilize the related knowledge for this task. We select a large corpus from a smart contract community Etherscan.io as our experimental subject. Extensive experimental results show the effectiveness of SCCLLM when compared with baselines in automatic evaluation and human evaluation.
[ { "created": "Fri, 17 Nov 2023 08:31:09 GMT", "version": "v1" }, { "created": "Tue, 16 Jan 2024 07:58:25 GMT", "version": "v2" } ]
2024-01-17
[ [ "Zhao", "Junjie", "" ], [ "Chen", "Xiang", "" ], [ "Yang", "Guang", "" ], [ "Shen", "Yiheng", "" ] ]
The previous smart contract code comment (SCC) generation approaches can be divided into two categories: fine-tuning paradigm-based approaches and information retrieval-based approaches. However, for the fine-tuning paradigm-based approaches, the performance may be limited by the quality of the gathered dataset for the downstream task and they may have knowledge-forgetting issues. While for the information retrieval-based approaches, it is difficult for them to generate high-quality comments if similar code does not exist in the historical repository. Therefore we want to utilize the domain knowledge related to SCC generation in large language models (LLMs) to alleviate the disadvantages of these two types of approaches. In this study, we propose an approach SCCLLM based on LLMs and in-context learning. Specifically, in the demonstration selection phase, SCCLLM retrieves the top-k code snippets from the historical corpus by considering syntax, semantics, and lexical information. In the in-context learning phase, SCCLLM utilizes the retrieved code snippets as demonstrations, which can help to utilize the related knowledge for this task. We select a large corpus from a smart contract community Etherscan.io as our experimental subject. Extensive experimental results show the effectiveness of SCCLLM when compared with baselines in automatic evaluation and human evaluation.
2108.10061
Larkin Liu
Larkin Liu, Jun Tao Luo
An Extensible and Modular Design and Implementation of Monte Carlo Tree Search for the JVM
18 pages, 7 figures, Manuscript
null
null
null
cs.LG stat.CO
http://creativecommons.org/licenses/by/4.0/
Flexible implementations of Monte Carlo Tree Search (MCTS), combined with domain specific knowledge and hybridization with other search algorithms, can be powerful for finding the solutions to problems in complex planning. We introduce mctreesearch4j, an MCTS implementation written as a standard JVM library following key design principles of object oriented programming. We define key class abstractions allowing the MCTS library to flexibly adapt to any well defined Markov Decision Process or turn-based adversarial game. Furthermore, our library is designed to be modular and extensible, utilizing class inheritance and generic typing to standardize custom algorithm definitions. We demonstrate that the design of the MCTS implementation provides ease of adaptation for unique heuristics and customization across varying Markov Decision Process (MDP) domains. In addition, the implementation is reasonably performant and accurate for standard MDP's. In addition, via the implementation of mctreesearch4j, the nuances of different types of MCTS algorithms are discussed.
[ { "created": "Fri, 30 Jul 2021 08:17:04 GMT", "version": "v1" } ]
2021-08-24
[ [ "Liu", "Larkin", "" ], [ "Luo", "Jun Tao", "" ] ]
Flexible implementations of Monte Carlo Tree Search (MCTS), combined with domain specific knowledge and hybridization with other search algorithms, can be powerful for finding the solutions to problems in complex planning. We introduce mctreesearch4j, an MCTS implementation written as a standard JVM library following key design principles of object oriented programming. We define key class abstractions allowing the MCTS library to flexibly adapt to any well defined Markov Decision Process or turn-based adversarial game. Furthermore, our library is designed to be modular and extensible, utilizing class inheritance and generic typing to standardize custom algorithm definitions. We demonstrate that the design of the MCTS implementation provides ease of adaptation for unique heuristics and customization across varying Markov Decision Process (MDP) domains. In addition, the implementation is reasonably performant and accurate for standard MDP's. In addition, via the implementation of mctreesearch4j, the nuances of different types of MCTS algorithms are discussed.
1606.03212
Furong Huang
Furong Huang
Discovery of Latent Factors in High-dimensional Data Using Tensor Methods
Ph.D. Thesis
null
null
null
cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Unsupervised learning aims at the discovery of hidden structure that drives the observations in the real world. It is essential for success in modern machine learning. Latent variable models are versatile in unsupervised learning and have applications in almost every domain. Training latent variable models is challenging due to the non-convexity of the likelihood objective. An alternative method is based on the spectral decomposition of low order moment tensors. This versatile framework is guaranteed to estimate the correct model consistently. My thesis spans both theoretical analysis of tensor decomposition framework and practical implementation of various applications. This thesis presents theoretical results on convergence to globally optimal solution of tensor decomposition using the stochastic gradient descent, despite non-convexity of the objective. This is the first work that gives global convergence guarantees for the stochastic gradient descent on non-convex functions with exponentially many local minima and saddle points. This thesis also presents large-scale deployment of spectral methods carried out on various platforms. Dimensionality reduction techniques such as random projection are incorporated for a highly parallel and scalable tensor decomposition algorithm. We obtain a gain in both accuracies and in running times by several orders of magnitude compared to the state-of-art variational methods. To solve real world problems, more advanced models and learning algorithms are proposed. This thesis discusses generalization of LDA model to mixed membership stochastic block model for learning user communities in social network, convolutional dictionary model for learning word-sequence embeddings, hierarchical tensor decomposition and latent tree structure model for learning disease hierarchy, and spatial point process mixture model for detecting cell types in neuroscience.
[ { "created": "Fri, 10 Jun 2016 07:17:00 GMT", "version": "v1" } ]
2016-06-13
[ [ "Huang", "Furong", "" ] ]
Unsupervised learning aims at the discovery of hidden structure that drives the observations in the real world. It is essential for success in modern machine learning. Latent variable models are versatile in unsupervised learning and have applications in almost every domain. Training latent variable models is challenging due to the non-convexity of the likelihood objective. An alternative method is based on the spectral decomposition of low order moment tensors. This versatile framework is guaranteed to estimate the correct model consistently. My thesis spans both theoretical analysis of tensor decomposition framework and practical implementation of various applications. This thesis presents theoretical results on convergence to globally optimal solution of tensor decomposition using the stochastic gradient descent, despite non-convexity of the objective. This is the first work that gives global convergence guarantees for the stochastic gradient descent on non-convex functions with exponentially many local minima and saddle points. This thesis also presents large-scale deployment of spectral methods carried out on various platforms. Dimensionality reduction techniques such as random projection are incorporated for a highly parallel and scalable tensor decomposition algorithm. We obtain a gain in both accuracies and in running times by several orders of magnitude compared to the state-of-art variational methods. To solve real world problems, more advanced models and learning algorithms are proposed. This thesis discusses generalization of LDA model to mixed membership stochastic block model for learning user communities in social network, convolutional dictionary model for learning word-sequence embeddings, hierarchical tensor decomposition and latent tree structure model for learning disease hierarchy, and spatial point process mixture model for detecting cell types in neuroscience.
2308.08991
Yuqiang Sun
Yuqiang Sun, Zhengzi Xu, Chengwei Liu, Yiran Zhang, Yang Liu
Who is the Real Hero? Measuring Developer Contribution via Multi-dimensional Data Integration
null
null
10.1109/ASE56229.2023.00102
null
cs.SE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Proper incentives are important for motivating developers in open-source communities, which is crucial for maintaining the development of open-source software healthy. To provide such incentives, an accurate and objective developer contribution measurement method is needed. However, existing methods rely heavily on manual peer review, lacking objectivity and transparency. The metrics of some automated works about effort estimation use only syntax-level or even text-level information, such as changed lines of code, which lack robustness. Furthermore, some works about identifying core developers provide only a qualitative understanding without a quantitative score or have some project-specific parameters, which makes them not practical in real-world projects. To this end, we propose CValue, a multidimensional information fusion-based approach to measure developer contributions. CValue extracts both syntax and semantic information from the source code changes in four dimensions: modification amount, understandability, inter-function and intra-function impact of modification. It fuses the information to produce the contribution score for each of the commits in the projects. Experimental results show that CValue outperforms other approaches by 19.59% on 10 real-world projects with manually labeled ground truth. We validated and proved that the performance of CValue, which takes 83.39 seconds per commit, is acceptable to be applied in real-world projects. Furthermore, we performed a large-scale experiment on 174 projects and detected 2,282 developers having inflated commits. Of these, 2,050 developers did not make any syntax contribution; and 103 were identified as bots.
[ { "created": "Thu, 17 Aug 2023 13:57:44 GMT", "version": "v1" }, { "created": "Thu, 31 Aug 2023 07:42:39 GMT", "version": "v2" } ]
2023-11-21
[ [ "Sun", "Yuqiang", "" ], [ "Xu", "Zhengzi", "" ], [ "Liu", "Chengwei", "" ], [ "Zhang", "Yiran", "" ], [ "Liu", "Yang", "" ] ]
Proper incentives are important for motivating developers in open-source communities, which is crucial for maintaining the development of open-source software healthy. To provide such incentives, an accurate and objective developer contribution measurement method is needed. However, existing methods rely heavily on manual peer review, lacking objectivity and transparency. The metrics of some automated works about effort estimation use only syntax-level or even text-level information, such as changed lines of code, which lack robustness. Furthermore, some works about identifying core developers provide only a qualitative understanding without a quantitative score or have some project-specific parameters, which makes them not practical in real-world projects. To this end, we propose CValue, a multidimensional information fusion-based approach to measure developer contributions. CValue extracts both syntax and semantic information from the source code changes in four dimensions: modification amount, understandability, inter-function and intra-function impact of modification. It fuses the information to produce the contribution score for each of the commits in the projects. Experimental results show that CValue outperforms other approaches by 19.59% on 10 real-world projects with manually labeled ground truth. We validated and proved that the performance of CValue, which takes 83.39 seconds per commit, is acceptable to be applied in real-world projects. Furthermore, we performed a large-scale experiment on 174 projects and detected 2,282 developers having inflated commits. Of these, 2,050 developers did not make any syntax contribution; and 103 were identified as bots.
1611.06385
Jop Bri\"et
Jop Bri\"et
On Embeddings of $\ell_1^k$ from Locally Decodable Codes
Appeared earlier on ECCC (http://eccc.hpi-web.de/report/2015/086/). This version has a slightly shorter abstract and slightly edited introduction. Removed left-over notes
null
null
null
cs.CC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We show that any $q$-query locally decodable code (LDC) gives a copy of $\ell_1^k$ with small distortion in the Banach space of $q$-linear forms on $\ell_{p_1}^N\times\cdots\times\ell_{p_q}^N$, provided $1/p_1 + \cdots + 1/p_q \leq 1$ and where $k$, $N$, and the distortion are simple functions of the code parameters. We exhibit the copy of $\ell_1^k$ by constructing a basis for it directly from "smooth" LDC decoders. Based on this, we give alternative proofs for known lower bounds on the length of 2-query LDCs. Using similar techniques, we reprove known lower bounds for larger $q$. We also discuss the relation with an alternative proof, due to Pisier, of a result of Naor, Regev, and the author on cotype properties of projective tensor products of $\ell_p$ spaces.
[ { "created": "Sat, 19 Nov 2016 15:39:20 GMT", "version": "v1" }, { "created": "Tue, 22 Nov 2016 12:04:37 GMT", "version": "v2" } ]
2016-11-23
[ [ "Briët", "Jop", "" ] ]
We show that any $q$-query locally decodable code (LDC) gives a copy of $\ell_1^k$ with small distortion in the Banach space of $q$-linear forms on $\ell_{p_1}^N\times\cdots\times\ell_{p_q}^N$, provided $1/p_1 + \cdots + 1/p_q \leq 1$ and where $k$, $N$, and the distortion are simple functions of the code parameters. We exhibit the copy of $\ell_1^k$ by constructing a basis for it directly from "smooth" LDC decoders. Based on this, we give alternative proofs for known lower bounds on the length of 2-query LDCs. Using similar techniques, we reprove known lower bounds for larger $q$. We also discuss the relation with an alternative proof, due to Pisier, of a result of Naor, Regev, and the author on cotype properties of projective tensor products of $\ell_p$ spaces.
1811.03966
Lars Jaffke
Lars Jaffke and Paloma T. Lima
A Complexity Dichotomy for Critical Values of the b-Chromatic Number of Graphs
20 pages, 1 figure
null
null
null
cs.DS cs.CC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
A $b$-coloring of a graph $G$ is a proper coloring of its vertices such that each color class contains a vertex that has at least one neighbor in all the other color classes. The b-Coloring problem asks whether a graph $G$ has a $b$-coloring with $k$ colors. The $b$-chromatic number of a graph $G$, denoted by $\chi_b(G)$, is the maximum number $k$ such that $G$ admits a $b$-coloring with $k$ colors. We consider the complexity of the b-Coloring problem, whenever the value of $k$ is close to one of two upper bounds on $\chi_b(G)$: The maximum degree $\Delta(G)$ plus one, and the $m$-degree, denoted by $m(G)$, which is defined as the maximum number $i$ such that $G$ has $i$ vertices of degree at least $i-1$. We obtain a dichotomy result stating that for fixed $k \in \{\Delta(G) + 1 - p, m(G) - p\}$, the problem is polynomial-time solvable whenever $p \in \{0, 1\}$ and, even when $k = 3$, it is NP-complete whenever $p \ge 2$. We furthermore consider parameterizations of the b-Coloring problem that involve the maximum degree $\Delta(G)$ of the input graph $G$ and give two FPT-algorithms. First, we show that deciding whether a graph $G$ has a $b$-coloring with $m(G)$ colors is FPT parameterized by $\Delta(G)$. Second, we show that b-Coloring is FPT parameterized by $\Delta(G) + \ell_k(G)$, where $\ell_k(G)$ denotes the number of vertices of degree at least $k$.
[ { "created": "Fri, 9 Nov 2018 15:22:35 GMT", "version": "v1" }, { "created": "Mon, 11 Feb 2019 15:04:17 GMT", "version": "v2" } ]
2019-02-12
[ [ "Jaffke", "Lars", "" ], [ "Lima", "Paloma T.", "" ] ]
A $b$-coloring of a graph $G$ is a proper coloring of its vertices such that each color class contains a vertex that has at least one neighbor in all the other color classes. The b-Coloring problem asks whether a graph $G$ has a $b$-coloring with $k$ colors. The $b$-chromatic number of a graph $G$, denoted by $\chi_b(G)$, is the maximum number $k$ such that $G$ admits a $b$-coloring with $k$ colors. We consider the complexity of the b-Coloring problem, whenever the value of $k$ is close to one of two upper bounds on $\chi_b(G)$: The maximum degree $\Delta(G)$ plus one, and the $m$-degree, denoted by $m(G)$, which is defined as the maximum number $i$ such that $G$ has $i$ vertices of degree at least $i-1$. We obtain a dichotomy result stating that for fixed $k \in \{\Delta(G) + 1 - p, m(G) - p\}$, the problem is polynomial-time solvable whenever $p \in \{0, 1\}$ and, even when $k = 3$, it is NP-complete whenever $p \ge 2$. We furthermore consider parameterizations of the b-Coloring problem that involve the maximum degree $\Delta(G)$ of the input graph $G$ and give two FPT-algorithms. First, we show that deciding whether a graph $G$ has a $b$-coloring with $m(G)$ colors is FPT parameterized by $\Delta(G)$. Second, we show that b-Coloring is FPT parameterized by $\Delta(G) + \ell_k(G)$, where $\ell_k(G)$ denotes the number of vertices of degree at least $k$.
1504.02089
Tomer Koren
Elad Hazan, Tomer Koren
The Computational Power of Optimization in Online Learning
null
null
null
null
cs.LG cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We consider the fundamental problem of prediction with expert advice where the experts are "optimizable": there is a black-box optimization oracle that can be used to compute, in constant time, the leading expert in retrospect at any point in time. In this setting, we give a novel online algorithm that attains vanishing regret with respect to $N$ experts in total $\widetilde{O}(\sqrt{N})$ computation time. We also give a lower bound showing that this running time cannot be improved (up to log factors) in the oracle model, thereby exhibiting a quadratic speedup as compared to the standard, oracle-free setting where the required time for vanishing regret is $\widetilde{\Theta}(N)$. These results demonstrate an exponential gap between the power of optimization in online learning and its power in statistical learning: in the latter, an optimization oracle---i.e., an efficient empirical risk minimizer---allows to learn a finite hypothesis class of size $N$ in time $O(\log{N})$. We also study the implications of our results to learning in repeated zero-sum games, in a setting where the players have access to oracles that compute, in constant time, their best-response to any mixed strategy of their opponent. We show that the runtime required for approximating the minimax value of the game in this setting is $\widetilde{\Theta}(\sqrt{N})$, yielding again a quadratic improvement upon the oracle-free setting, where $\widetilde{\Theta}(N)$ is known to be tight.
[ { "created": "Wed, 8 Apr 2015 19:54:27 GMT", "version": "v1" }, { "created": "Fri, 17 Apr 2015 12:15:37 GMT", "version": "v2" }, { "created": "Mon, 2 Nov 2015 20:30:50 GMT", "version": "v3" }, { "created": "Wed, 27 Jan 2016 09:07:59 GMT", "version": "v4" } ]
2016-01-28
[ [ "Hazan", "Elad", "" ], [ "Koren", "Tomer", "" ] ]
We consider the fundamental problem of prediction with expert advice where the experts are "optimizable": there is a black-box optimization oracle that can be used to compute, in constant time, the leading expert in retrospect at any point in time. In this setting, we give a novel online algorithm that attains vanishing regret with respect to $N$ experts in total $\widetilde{O}(\sqrt{N})$ computation time. We also give a lower bound showing that this running time cannot be improved (up to log factors) in the oracle model, thereby exhibiting a quadratic speedup as compared to the standard, oracle-free setting where the required time for vanishing regret is $\widetilde{\Theta}(N)$. These results demonstrate an exponential gap between the power of optimization in online learning and its power in statistical learning: in the latter, an optimization oracle---i.e., an efficient empirical risk minimizer---allows to learn a finite hypothesis class of size $N$ in time $O(\log{N})$. We also study the implications of our results to learning in repeated zero-sum games, in a setting where the players have access to oracles that compute, in constant time, their best-response to any mixed strategy of their opponent. We show that the runtime required for approximating the minimax value of the game in this setting is $\widetilde{\Theta}(\sqrt{N})$, yielding again a quadratic improvement upon the oracle-free setting, where $\widetilde{\Theta}(N)$ is known to be tight.
1701.01170
Yangzihao Wang
Yangzihao Wang, Yuechao Pan, Andrew Davidson, Yuduo Wu, Carl Yang, Leyuan Wang, Muhammad Osama, Chenshan Yuan, Weitang Liu, Andy T. Riffel and John D. Owens
Gunrock: GPU Graph Analytics
52 pages, invited paper to ACM Transactions on Parallel Computing (TOPC), an extended version of PPoPP'16 paper "Gunrock: A High-Performance Graph Processing Library on the GPU"
null
null
null
cs.DC
http://creativecommons.org/licenses/by/4.0/
For large-scale graph analytics on the GPU, the irregularity of data access and control flow, and the complexity of programming GPUs, have presented two significant challenges to developing a programmable high-performance graph library. "Gunrock", our graph-processing system designed specifically for the GPU, uses a high-level, bulk-synchronous, data-centric abstraction focused on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high performance GPU computing primitives and optimization strategies with a high-level programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We characterize the performance of various optimization strategies and evaluate Gunrock's overall performance on different GPU architectures on a wide range of graph primitives that span from traversal-based algorithms and ranking algorithms, to triangle counting and bipartite-graph-based algorithms. The results show that on a single GPU, Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives and CPU shared-memory graph libraries such as Ligra and Galois, and better performance than any other GPU high-level graph library.
[ { "created": "Wed, 4 Jan 2017 22:16:07 GMT", "version": "v1" } ]
2017-01-06
[ [ "Wang", "Yangzihao", "" ], [ "Pan", "Yuechao", "" ], [ "Davidson", "Andrew", "" ], [ "Wu", "Yuduo", "" ], [ "Yang", "Carl", "" ], [ "Wang", "Leyuan", "" ], [ "Osama", "Muhammad", "" ], [ "Yuan", "Chenshan", "" ], [ "Liu", "Weitang", "" ], [ "Riffel", "Andy T.", "" ], [ "Owens", "John D.", "" ] ]
For large-scale graph analytics on the GPU, the irregularity of data access and control flow, and the complexity of programming GPUs, have presented two significant challenges to developing a programmable high-performance graph library. "Gunrock", our graph-processing system designed specifically for the GPU, uses a high-level, bulk-synchronous, data-centric abstraction focused on operations on a vertex or edge frontier. Gunrock achieves a balance between performance and expressiveness by coupling high performance GPU computing primitives and optimization strategies with a high-level programming model that allows programmers to quickly develop new graph primitives with small code size and minimal GPU programming knowledge. We characterize the performance of various optimization strategies and evaluate Gunrock's overall performance on different GPU architectures on a wide range of graph primitives that span from traversal-based algorithms and ranking algorithms, to triangle counting and bipartite-graph-based algorithms. The results show that on a single GPU, Gunrock has on average at least an order of magnitude speedup over Boost and PowerGraph, comparable performance to the fastest GPU hardwired primitives and CPU shared-memory graph libraries such as Ligra and Galois, and better performance than any other GPU high-level graph library.
2208.04022
Wen-Ji Zhou
Qianying Lin, Wen-Ji Zhou, Yanshi Wang, Qing Da, Qing-Guo Chen, Bing Wang
Sparse Attentive Memory Network for Click-through Rate Prediction with Long Sequences
Published as a conference paper at the 31st ACM International Conference on Information and Knowledge Management, CIKM 2022
null
10.1145/3511808.3557095
null
cs.IR cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sequential recommendation predicts users' next behaviors with their historical interactions. Recommending with longer sequences improves recommendation accuracy and increases the degree of personalization. As sequences get longer, existing works have not yet addressed the following two main challenges. Firstly, modeling long-range intra-sequence dependency is difficult with increasing sequence lengths. Secondly, it requires efficient memory and computational speeds. In this paper, we propose a Sparse Attentive Memory (SAM) network for long sequential user behavior modeling. SAM supports efficient training and real-time inference for user behavior sequences with lengths on the scale of thousands. In SAM, we model the target item as the query and the long sequence as the knowledge database, where the former continuously elicits relevant information from the latter. SAM simultaneously models target-sequence dependencies and long-range intra-sequence dependencies with O(L) complexity and O(1) number of sequential updates, which can only be achieved by the self-attention mechanism with O(L^2) complexity. Extensive empirical results demonstrate that our proposed solution is effective not only in long user behavior modeling but also on short sequences modeling. Implemented on sequences of length 1000, SAM is successfully deployed on one of the largest international E-commerce platforms. This inference time is within 30ms, with a substantial 7.30% click-through rate improvement for the online A/B test. To the best of our knowledge, it is the first end-to-end long user sequence modeling framework that models intra-sequence and target-sequence dependencies with the aforementioned degree of efficiency and successfully deployed on a large-scale real-time industrial recommender system.
[ { "created": "Mon, 8 Aug 2022 10:11:46 GMT", "version": "v1" }, { "created": "Fri, 2 Sep 2022 06:17:49 GMT", "version": "v2" } ]
2022-09-05
[ [ "Lin", "Qianying", "" ], [ "Zhou", "Wen-Ji", "" ], [ "Wang", "Yanshi", "" ], [ "Da", "Qing", "" ], [ "Chen", "Qing-Guo", "" ], [ "Wang", "Bing", "" ] ]
Sequential recommendation predicts users' next behaviors with their historical interactions. Recommending with longer sequences improves recommendation accuracy and increases the degree of personalization. As sequences get longer, existing works have not yet addressed the following two main challenges. Firstly, modeling long-range intra-sequence dependency is difficult with increasing sequence lengths. Secondly, it requires efficient memory and computational speeds. In this paper, we propose a Sparse Attentive Memory (SAM) network for long sequential user behavior modeling. SAM supports efficient training and real-time inference for user behavior sequences with lengths on the scale of thousands. In SAM, we model the target item as the query and the long sequence as the knowledge database, where the former continuously elicits relevant information from the latter. SAM simultaneously models target-sequence dependencies and long-range intra-sequence dependencies with O(L) complexity and O(1) number of sequential updates, which can only be achieved by the self-attention mechanism with O(L^2) complexity. Extensive empirical results demonstrate that our proposed solution is effective not only in long user behavior modeling but also on short sequences modeling. Implemented on sequences of length 1000, SAM is successfully deployed on one of the largest international E-commerce platforms. This inference time is within 30ms, with a substantial 7.30% click-through rate improvement for the online A/B test. To the best of our knowledge, it is the first end-to-end long user sequence modeling framework that models intra-sequence and target-sequence dependencies with the aforementioned degree of efficiency and successfully deployed on a large-scale real-time industrial recommender system.
2208.14571
Zhen Zhang
Zhen Zhang, Ignavier Ng, Dong Gong, Yuhang Liu, Ehsan M Abbasnejad, Mingming Gong, Kun Zhang, Javen Qinfeng Shi
Truncated Matrix Power Iteration for Differentiable DAG Learning
Published in NeurIPS 2022
null
null
null
cs.LG cs.AI stat.ML
http://creativecommons.org/licenses/by-nc-sa/4.0/
Recovering underlying Directed Acyclic Graph (DAG) structures from observational data is highly challenging due to the combinatorial nature of the DAG-constrained optimization problem. Recently, DAG learning has been cast as a continuous optimization problem by characterizing the DAG constraint as a smooth equality one, generally based on polynomials over adjacency matrices. Existing methods place very small coefficients on high-order polynomial terms for stabilization, since they argue that large coefficients on the higher-order terms are harmful due to numeric exploding. On the contrary, we discover that large coefficients on higher-order terms are beneficial for DAG learning, when the spectral radiuses of the adjacency matrices are small, and that larger coefficients for higher-order terms can approximate the DAG constraints much better than the small counterparts. Based on this, we propose a novel DAG learning method with efficient truncated matrix power iteration to approximate geometric series based DAG constraints. Empirically, our DAG learning method outperforms the previous state-of-the-arts in various settings, often by a factor of $3$ or more in terms of structural Hamming distance.
[ { "created": "Tue, 30 Aug 2022 23:56:12 GMT", "version": "v1" }, { "created": "Wed, 21 Dec 2022 03:21:04 GMT", "version": "v2" } ]
2022-12-23
[ [ "Zhang", "Zhen", "" ], [ "Ng", "Ignavier", "" ], [ "Gong", "Dong", "" ], [ "Liu", "Yuhang", "" ], [ "Abbasnejad", "Ehsan M", "" ], [ "Gong", "Mingming", "" ], [ "Zhang", "Kun", "" ], [ "Shi", "Javen Qinfeng", "" ] ]
Recovering underlying Directed Acyclic Graph (DAG) structures from observational data is highly challenging due to the combinatorial nature of the DAG-constrained optimization problem. Recently, DAG learning has been cast as a continuous optimization problem by characterizing the DAG constraint as a smooth equality one, generally based on polynomials over adjacency matrices. Existing methods place very small coefficients on high-order polynomial terms for stabilization, since they argue that large coefficients on the higher-order terms are harmful due to numeric exploding. On the contrary, we discover that large coefficients on higher-order terms are beneficial for DAG learning, when the spectral radiuses of the adjacency matrices are small, and that larger coefficients for higher-order terms can approximate the DAG constraints much better than the small counterparts. Based on this, we propose a novel DAG learning method with efficient truncated matrix power iteration to approximate geometric series based DAG constraints. Empirically, our DAG learning method outperforms the previous state-of-the-arts in various settings, often by a factor of $3$ or more in terms of structural Hamming distance.
2011.06691
Fabien Racape
Franck Galpin, Fabien Racap\'e, Sunil Jaiswal, Philippe Bordes, Fabrice Le L\'eannec, Edouard Fran\c{c}ois
CNN-based driving of block partitioning for intra slices encoding
10 pages
2019 Data Compression Conference (DCC)
10.1109/DCC.2019.00024
null
cs.MM
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper provides a technical overview of a deep-learning-based encoder method aiming at optimizing next generation hybrid video encoders for driving the block partitioning in intra slices. An encoding approach based on Convolutional Neural Networks is explored to partly substitute classical heuristics-based encoder speed-ups by a systematic and automatic process. The solution allows controlling the trade-off between complexity and coding gains, in intra slices, with one single parameter. This algorithm was proposed at the Call for Proposals of the Joint Video Exploration Team (JVET) on video compression with capability beyond HEVC. In All Intra configuration, for a given allowed topology of splits, a speed-up of $\times 2$ is obtained without BD-rate loss, or a speed-up above $\times 4$ with a loss below 1\% in BD-rate.
[ { "created": "Thu, 12 Nov 2020 23:55:12 GMT", "version": "v1" } ]
2020-11-16
[ [ "Galpin", "Franck", "" ], [ "Racapé", "Fabien", "" ], [ "Jaiswal", "Sunil", "" ], [ "Bordes", "Philippe", "" ], [ "Léannec", "Fabrice Le", "" ], [ "François", "Edouard", "" ] ]
This paper provides a technical overview of a deep-learning-based encoder method aiming at optimizing next generation hybrid video encoders for driving the block partitioning in intra slices. An encoding approach based on Convolutional Neural Networks is explored to partly substitute classical heuristics-based encoder speed-ups by a systematic and automatic process. The solution allows controlling the trade-off between complexity and coding gains, in intra slices, with one single parameter. This algorithm was proposed at the Call for Proposals of the Joint Video Exploration Team (JVET) on video compression with capability beyond HEVC. In All Intra configuration, for a given allowed topology of splits, a speed-up of $\times 2$ is obtained without BD-rate loss, or a speed-up above $\times 4$ with a loss below 1\% in BD-rate.
1605.04478
Hamid Tizhoosh
Mina Nouredanesh, Hamid R. Tizhoosh, Ershad Banijamali
Gabor Barcodes for Medical Image Retrieval
To appear in proceedings of The 2016 IEEE International Conference on Image Processing (ICIP 2016), Sep 25-28, 2016, Phoenix, Arizona, USA
null
10.1109/ICIP.2016.7532807
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In recent years, advances in medical imaging have led to the emergence of massive databases, containing images from a diverse range of modalities. This has significantly heightened the need for automated annotation of the images on one side, and fast and memory-efficient content-based image retrieval systems on the other side. Binary descriptors have recently gained more attention as a potential vehicle to achieve these goals. One of the recently introduced binary descriptors for tagging of medical images are Radon barcodes (RBCs) that are driven from Radon transform via local thresholding. Gabor transform is also a powerful transform to extract texture-based information. Gabor features have exhibited robustness against rotation, scale, and also photometric disturbances, such as illumination changes and image noise in many applications. This paper introduces Gabor Barcodes (GBCs), as a novel framework for the image annotation. To find the most discriminative GBC for a given query image, the effects of employing Gabor filters with different parameters, i.e., different sets of scales and orientations, are investigated, resulting in different barcode lengths and retrieval performances. The proposed method has been evaluated on the IRMA dataset with 193 classes comprising of 12,677 x-ray images for indexing, and 1,733 x-rays images for testing. A total error score as low as $351$ ($\approx 80\%$ accuracy for the first hit) was achieved.
[ { "created": "Sat, 14 May 2016 22:39:29 GMT", "version": "v1" } ]
2016-11-15
[ [ "Nouredanesh", "Mina", "" ], [ "Tizhoosh", "Hamid R.", "" ], [ "Banijamali", "Ershad", "" ] ]
In recent years, advances in medical imaging have led to the emergence of massive databases, containing images from a diverse range of modalities. This has significantly heightened the need for automated annotation of the images on one side, and fast and memory-efficient content-based image retrieval systems on the other side. Binary descriptors have recently gained more attention as a potential vehicle to achieve these goals. One of the recently introduced binary descriptors for tagging of medical images are Radon barcodes (RBCs) that are driven from Radon transform via local thresholding. Gabor transform is also a powerful transform to extract texture-based information. Gabor features have exhibited robustness against rotation, scale, and also photometric disturbances, such as illumination changes and image noise in many applications. This paper introduces Gabor Barcodes (GBCs), as a novel framework for the image annotation. To find the most discriminative GBC for a given query image, the effects of employing Gabor filters with different parameters, i.e., different sets of scales and orientations, are investigated, resulting in different barcode lengths and retrieval performances. The proposed method has been evaluated on the IRMA dataset with 193 classes comprising of 12,677 x-ray images for indexing, and 1,733 x-rays images for testing. A total error score as low as $351$ ($\approx 80\%$ accuracy for the first hit) was achieved.
2108.08768
Abdullatif Albaseer Mr
Abdullatif Albaseer, Mohamed Abdallah, Ala Al-Fuqaha, and Aiman Erbad
Client Selection Approach in Support of Clustered Federated Learning over Wireless Edge Networks
4 figures, 7 pages
null
null
null
cs.DC cs.LG
http://creativecommons.org/licenses/by/4.0/
Clustered Federated Multitask Learning (CFL) was introduced as an efficient scheme to obtain reliable specialized models when data is imbalanced and distributed in a non-i.i.d. (non-independent and identically distributed) fashion amongst clients. While a similarity measure metric, like the cosine similarity, can be used to endow groups of the client with a specialized model, this process can be arduous as the server should involve all clients in each of the federated learning rounds. Therefore, it is imperative that a subset of clients is selected periodically due to the limited bandwidth and latency constraints at the network edge. To this end, this paper proposes a new client selection algorithm that aims to accelerate the convergence rate for obtaining specialized machine learning models that achieve high test accuracies for all client groups. Specifically, we introduce a client selection approach that leverages the devices' heterogeneity to schedule the clients based on their round latency and exploits the bandwidth reuse for clients that consume more time to update the model. Then, the server performs model averaging and clusters the clients based on predefined thresholds. When a specific cluster reaches a stationary point, the proposed algorithm uses a greedy scheduling algorithm for that group by selecting the clients with less latency to update the model. Extensive experiments show that the proposed approach lowers the training time and accelerates the convergence rate by up to 50% while imbuing each client with a specialized model that is fit for its local data distribution.
[ { "created": "Mon, 16 Aug 2021 21:38:22 GMT", "version": "v1" } ]
2021-08-20
[ [ "Albaseer", "Abdullatif", "" ], [ "Abdallah", "Mohamed", "" ], [ "Al-Fuqaha", "Ala", "" ], [ "Erbad", "Aiman", "" ] ]
Clustered Federated Multitask Learning (CFL) was introduced as an efficient scheme to obtain reliable specialized models when data is imbalanced and distributed in a non-i.i.d. (non-independent and identically distributed) fashion amongst clients. While a similarity measure metric, like the cosine similarity, can be used to endow groups of the client with a specialized model, this process can be arduous as the server should involve all clients in each of the federated learning rounds. Therefore, it is imperative that a subset of clients is selected periodically due to the limited bandwidth and latency constraints at the network edge. To this end, this paper proposes a new client selection algorithm that aims to accelerate the convergence rate for obtaining specialized machine learning models that achieve high test accuracies for all client groups. Specifically, we introduce a client selection approach that leverages the devices' heterogeneity to schedule the clients based on their round latency and exploits the bandwidth reuse for clients that consume more time to update the model. Then, the server performs model averaging and clusters the clients based on predefined thresholds. When a specific cluster reaches a stationary point, the proposed algorithm uses a greedy scheduling algorithm for that group by selecting the clients with less latency to update the model. Extensive experiments show that the proposed approach lowers the training time and accelerates the convergence rate by up to 50% while imbuing each client with a specialized model that is fit for its local data distribution.
2101.09409
Shin-Cheng Mu
Shin-Cheng Mu
Calculating a backtracking algorithm: an exercise in monadic program derivation
null
null
null
TR-IIS-19-003, Institute of Information Science, Academia Sinica
cs.PL
http://creativecommons.org/licenses/by-sa/4.0/
Equational reasoning is among the most important tools that functional programming provides us. Curiously, relatively less attention has been paid to reasoning about monadic programs. In this report we derive a backtracking algorithm for problem specifications that use a monadic unfold to generate possible solutions, which are filtered using a $\mathit{scanl}$-like predicate. We develop theorems that convert a variation of $\mathit{scanl}$ to a $\mathit{foldr}$ that uses the state monad, as well as theorems constructing hylomorphism. The algorithm is used to solve the $n$-queens puzzle, our running example. The aim is to develop theorems and patterns useful for the derivation of monadic programs, focusing on the intricate interaction between state and non-determinism.
[ { "created": "Sat, 23 Jan 2021 03:27:20 GMT", "version": "v1" } ]
2021-01-26
[ [ "Mu", "Shin-Cheng", "" ] ]
Equational reasoning is among the most important tools that functional programming provides us. Curiously, relatively less attention has been paid to reasoning about monadic programs. In this report we derive a backtracking algorithm for problem specifications that use a monadic unfold to generate possible solutions, which are filtered using a $\mathit{scanl}$-like predicate. We develop theorems that convert a variation of $\mathit{scanl}$ to a $\mathit{foldr}$ that uses the state monad, as well as theorems constructing hylomorphism. The algorithm is used to solve the $n$-queens puzzle, our running example. The aim is to develop theorems and patterns useful for the derivation of monadic programs, focusing on the intricate interaction between state and non-determinism.
2004.09141
Jimmy Wu
Jimmy Wu, Xingyuan Sun, Andy Zeng, Shuran Song, Johnny Lee, Szymon Rusinkiewicz, Thomas Funkhouser
Spatial Action Maps for Mobile Manipulation
To appear at Robotics: Science and Systems (RSS), 2020. Project page: https://spatial-action-maps.cs.princeton.edu
null
10.15607/RSS.2020.XVI.035
null
cs.RO cs.AI cs.CV cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Typical end-to-end formulations for learning robotic navigation involve predicting a small set of steering command actions (e.g., step forward, turn left, turn right, etc.) from images of the current state (e.g., a bird's-eye view of a SLAM reconstruction). Instead, we show that it can be advantageous to learn with dense action representations defined in the same domain as the state. In this work, we present "spatial action maps," in which the set of possible actions is represented by a pixel map (aligned with the input image of the current state), where each pixel represents a local navigational endpoint at the corresponding scene location. Using ConvNets to infer spatial action maps from state images, action predictions are thereby spatially anchored on local visual features in the scene, enabling significantly faster learning of complex behaviors for mobile manipulation tasks with reinforcement learning. In our experiments, we task a robot with pushing objects to a goal location, and find that policies learned with spatial action maps achieve much better performance than traditional alternatives.
[ { "created": "Mon, 20 Apr 2020 09:06:10 GMT", "version": "v1" }, { "created": "Thu, 4 Jun 2020 10:56:49 GMT", "version": "v2" } ]
2020-10-13
[ [ "Wu", "Jimmy", "" ], [ "Sun", "Xingyuan", "" ], [ "Zeng", "Andy", "" ], [ "Song", "Shuran", "" ], [ "Lee", "Johnny", "" ], [ "Rusinkiewicz", "Szymon", "" ], [ "Funkhouser", "Thomas", "" ] ]
Typical end-to-end formulations for learning robotic navigation involve predicting a small set of steering command actions (e.g., step forward, turn left, turn right, etc.) from images of the current state (e.g., a bird's-eye view of a SLAM reconstruction). Instead, we show that it can be advantageous to learn with dense action representations defined in the same domain as the state. In this work, we present "spatial action maps," in which the set of possible actions is represented by a pixel map (aligned with the input image of the current state), where each pixel represents a local navigational endpoint at the corresponding scene location. Using ConvNets to infer spatial action maps from state images, action predictions are thereby spatially anchored on local visual features in the scene, enabling significantly faster learning of complex behaviors for mobile manipulation tasks with reinforcement learning. In our experiments, we task a robot with pushing objects to a goal location, and find that policies learned with spatial action maps achieve much better performance than traditional alternatives.
2101.06398
JianYu Wang
Jianyu Wang, Shanzheng Guan, Shupei Liu, Xiao-Lei Zhang
Minimum-volume Multichannel Nonnegative matrix factorization for blind source separation
null
null
null
null
cs.SD eess.AS
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Multichannel blind audio source separation aims to recover the latent sources from their multichannel mixtures without supervised information. One state-of-the-art blind audio source separation method, named independent low-rank matrix analysis (ILRMA), unifies independent vector analysis (IVA) and nonnegative matrix factorization (NMF). However, the spectra matrix produced from NMF may not find a compact spectral basis. It may not guarantee the identifiability of each source as well. To address this problem, here we propose to enhance the identifiability of the source model by a minimum-volume prior distribution. We further regularize a multichannel NMF (MNMF) and ILRMA respectively with the minimum-volume regularizer. The proposed methods maximize the posterior distribution of the separated sources, which ensures the stability of the convergence. Experimental results demonstrate the effectiveness of the proposed methods compared with auxiliary independent vector analysis, MNMF, ILRMA and its extensions.
[ { "created": "Sat, 16 Jan 2021 08:12:23 GMT", "version": "v1" }, { "created": "Tue, 30 Mar 2021 03:11:58 GMT", "version": "v2" } ]
2021-03-31
[ [ "Wang", "Jianyu", "" ], [ "Guan", "Shanzheng", "" ], [ "Liu", "Shupei", "" ], [ "Zhang", "Xiao-Lei", "" ] ]
Multichannel blind audio source separation aims to recover the latent sources from their multichannel mixtures without supervised information. One state-of-the-art blind audio source separation method, named independent low-rank matrix analysis (ILRMA), unifies independent vector analysis (IVA) and nonnegative matrix factorization (NMF). However, the spectra matrix produced from NMF may not find a compact spectral basis. It may not guarantee the identifiability of each source as well. To address this problem, here we propose to enhance the identifiability of the source model by a minimum-volume prior distribution. We further regularize a multichannel NMF (MNMF) and ILRMA respectively with the minimum-volume regularizer. The proposed methods maximize the posterior distribution of the separated sources, which ensures the stability of the convergence. Experimental results demonstrate the effectiveness of the proposed methods compared with auxiliary independent vector analysis, MNMF, ILRMA and its extensions.
2104.09176
Stefan Zernetsch
Stefan Zernetsch, Hannes Reichert, Viktor Kress, Konrad Doll, Bernhard Sick
Cyclist Intention Detection: A Probabilistic Approach
null
null
null
null
cs.CV cs.LG
http://creativecommons.org/licenses/by/4.0/
This article presents a holistic approach for probabilistic cyclist intention detection. A basic movement detection based on motion history images (MHI) and a residual convolutional neural network (ResNet) are used to estimate probabilities for the current cyclist motion state. These probabilities are used as weights in a probabilistic ensemble trajectory forecast. The ensemble consists of specialized models, which produce individual forecasts in the form of Gaussian distributions under the assumption of a certain motion state of the cyclist (e.g. cyclist is starting or turning left). By weighting the specialized models, we create forecasts in the from of Gaussian mixtures that define regions within which the cyclists will reside with a certain probability. To evaluate our method, we rate the reliability, sharpness, and positional accuracy of our forecasted distributions. We compare our method to a single model approach which produces forecasts in the form of Gaussian distributions and show that our method is able to produce more reliable and sharper outputs while retaining comparable positional accuracy. Both methods are evaluated using a dataset created at a public traffic intersection. Our code and the dataset are made publicly available.
[ { "created": "Mon, 19 Apr 2021 09:59:04 GMT", "version": "v1" } ]
2021-04-20
[ [ "Zernetsch", "Stefan", "" ], [ "Reichert", "Hannes", "" ], [ "Kress", "Viktor", "" ], [ "Doll", "Konrad", "" ], [ "Sick", "Bernhard", "" ] ]
This article presents a holistic approach for probabilistic cyclist intention detection. A basic movement detection based on motion history images (MHI) and a residual convolutional neural network (ResNet) are used to estimate probabilities for the current cyclist motion state. These probabilities are used as weights in a probabilistic ensemble trajectory forecast. The ensemble consists of specialized models, which produce individual forecasts in the form of Gaussian distributions under the assumption of a certain motion state of the cyclist (e.g. cyclist is starting or turning left). By weighting the specialized models, we create forecasts in the from of Gaussian mixtures that define regions within which the cyclists will reside with a certain probability. To evaluate our method, we rate the reliability, sharpness, and positional accuracy of our forecasted distributions. We compare our method to a single model approach which produces forecasts in the form of Gaussian distributions and show that our method is able to produce more reliable and sharper outputs while retaining comparable positional accuracy. Both methods are evaluated using a dataset created at a public traffic intersection. Our code and the dataset are made publicly available.
2105.04112
Jun Yang
Jun Yang, Yizhou Gao, Dong Li, Steven L. Waslander
ROBI: A Multi-View Dataset for Reflective Objects in Robotic Bin-Picking
null
null
null
null
cs.RO
http://creativecommons.org/licenses/by-nc-sa/4.0/
In robotic bin-picking applications, the perception of texture-less, highly reflective parts is a valuable but challenging task. The high glossiness can introduce fake edges in RGB images and inaccurate depth measurements especially in heavily cluttered bin scenario. In this paper, we present the ROBI (Reflective Objects in BIns) dataset, a public dataset for 6D object pose estimation and multi-view depth fusion in robotic bin-picking scenarios. The ROBI dataset includes a total of 63 bin-picking scenes captured with two active stereo camera: a high-cost Ensenso sensor and a low-cost RealSense sensor. For each scene, the monochrome/RGB images and depth maps are captured from sampled view spheres around the scene, and are annotated with accurate 6D poses of visible objects and an associated visibility score. For evaluating the performance of depth fusion, we captured the ground truth depth maps by high-cost Ensenso camera with objects coated in anti-reflective scanning spray. To show the utility of the dataset, we evaluated the representative algorithms of 6D object pose estimation and multi-view depth fusion on the full dataset. Evaluation results demonstrate the difficulty of highly reflective objects, especially in difficult cases due to the degradation of depth data quality, severe occlusions and cluttered scene. The ROBI dataset is available online at https://www.trailab.utias.utoronto.ca/robi.
[ { "created": "Mon, 10 May 2021 04:55:29 GMT", "version": "v1" }, { "created": "Thu, 7 Oct 2021 01:12:20 GMT", "version": "v2" } ]
2021-10-08
[ [ "Yang", "Jun", "" ], [ "Gao", "Yizhou", "" ], [ "Li", "Dong", "" ], [ "Waslander", "Steven L.", "" ] ]
In robotic bin-picking applications, the perception of texture-less, highly reflective parts is a valuable but challenging task. The high glossiness can introduce fake edges in RGB images and inaccurate depth measurements especially in heavily cluttered bin scenario. In this paper, we present the ROBI (Reflective Objects in BIns) dataset, a public dataset for 6D object pose estimation and multi-view depth fusion in robotic bin-picking scenarios. The ROBI dataset includes a total of 63 bin-picking scenes captured with two active stereo camera: a high-cost Ensenso sensor and a low-cost RealSense sensor. For each scene, the monochrome/RGB images and depth maps are captured from sampled view spheres around the scene, and are annotated with accurate 6D poses of visible objects and an associated visibility score. For evaluating the performance of depth fusion, we captured the ground truth depth maps by high-cost Ensenso camera with objects coated in anti-reflective scanning spray. To show the utility of the dataset, we evaluated the representative algorithms of 6D object pose estimation and multi-view depth fusion on the full dataset. Evaluation results demonstrate the difficulty of highly reflective objects, especially in difficult cases due to the degradation of depth data quality, severe occlusions and cluttered scene. The ROBI dataset is available online at https://www.trailab.utias.utoronto.ca/robi.
2311.06728
Xiyue Gao
Xiyue Gao, Zhuang Liu, Jiangtao Cui, Hui Li, Hui Zhang, Kewei Wei, Kankan Zhao
A Comprehensive Survey on Database Management System Fuzzing: Techniques, Taxonomy and Experimental Comparison
34 pages, 22 figures
null
null
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Database Management System (DBMS) fuzzing is an automated testing technique aimed at detecting errors and vulnerabilities in DBMSs by generating, mutating, and executing test cases. It not only reduces the time and cost of manual testing but also enhances detection coverage, providing valuable assistance in developing commercial DBMSs. Existing fuzzing surveys mainly focus on general-purpose software. However, DBMSs are different from them in terms of internal structure, input/output, and test objectives, requiring specialized fuzzing strategies. Therefore, this paper focuses on DBMS fuzzing and provides a comprehensive review and comparison of the methods in this field. We first introduce the fundamental concepts. Then, we systematically define a general fuzzing procedure and decompose and categorize existing methods. Furthermore, we classify existing methods from the testing objective perspective, covering various components in DBMSs. For representative works, more detailed descriptions are provided to analyze their strengths and limitations. To objectively evaluate the performance of each method, we present an open-source DBMS fuzzing toolkit, OpenDBFuzz. Based on this toolkit, we conduct a detailed experimental comparative analysis of existing methods and finally discuss future research directions.
[ { "created": "Sun, 12 Nov 2023 04:18:03 GMT", "version": "v1" } ]
2023-11-14
[ [ "Gao", "Xiyue", "" ], [ "Liu", "Zhuang", "" ], [ "Cui", "Jiangtao", "" ], [ "Li", "Hui", "" ], [ "Zhang", "Hui", "" ], [ "Wei", "Kewei", "" ], [ "Zhao", "Kankan", "" ] ]
Database Management System (DBMS) fuzzing is an automated testing technique aimed at detecting errors and vulnerabilities in DBMSs by generating, mutating, and executing test cases. It not only reduces the time and cost of manual testing but also enhances detection coverage, providing valuable assistance in developing commercial DBMSs. Existing fuzzing surveys mainly focus on general-purpose software. However, DBMSs are different from them in terms of internal structure, input/output, and test objectives, requiring specialized fuzzing strategies. Therefore, this paper focuses on DBMS fuzzing and provides a comprehensive review and comparison of the methods in this field. We first introduce the fundamental concepts. Then, we systematically define a general fuzzing procedure and decompose and categorize existing methods. Furthermore, we classify existing methods from the testing objective perspective, covering various components in DBMSs. For representative works, more detailed descriptions are provided to analyze their strengths and limitations. To objectively evaluate the performance of each method, we present an open-source DBMS fuzzing toolkit, OpenDBFuzz. Based on this toolkit, we conduct a detailed experimental comparative analysis of existing methods and finally discuss future research directions.
2401.09410
Bhaskar Mitra
Karina Corti\~nas-Lorenzo, Si\^an Lindley, Ida Larsen-Ledet and Bhaskar Mitra
Through the Looking-Glass: Transparency Implications and Challenges in Enterprise AI Knowledge Systems
null
null
null
null
cs.CY cs.AI cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Knowledge can't be disentangled from people. As AI knowledge systems mine vast volumes of work-related data, the knowledge that's being extracted and surfaced is intrinsically linked to the people who create and use it. When these systems get embedded in organizational settings, the information that is brought to the foreground and the information that's pushed to the periphery can influence how individuals see each other and how they see themselves at work. In this paper, we present the looking-glass metaphor and use it to conceptualize AI knowledge systems as systems that reflect and distort, expanding our view on transparency requirements, implications and challenges. We formulate transparency as a key mediator in shaping different ways of seeing, including seeing into the system, which unveils its capabilities, limitations and behavior, and seeing through the system, which shapes workers' perceptions of their own contributions and others within the organization. Recognizing the sociotechnical nature of these systems, we identify three transparency dimensions necessary to realize the value of AI knowledge systems, namely system transparency, procedural transparency and transparency of outcomes. We discuss key challenges hindering the implementation of these forms of transparency, bringing to light the wider sociotechnical gap and highlighting directions for future Computer-supported Cooperative Work (CSCW) research.
[ { "created": "Wed, 17 Jan 2024 18:47:30 GMT", "version": "v1" } ]
2024-01-18
[ [ "Cortiñas-Lorenzo", "Karina", "" ], [ "Lindley", "Siân", "" ], [ "Larsen-Ledet", "Ida", "" ], [ "Mitra", "Bhaskar", "" ] ]
Knowledge can't be disentangled from people. As AI knowledge systems mine vast volumes of work-related data, the knowledge that's being extracted and surfaced is intrinsically linked to the people who create and use it. When these systems get embedded in organizational settings, the information that is brought to the foreground and the information that's pushed to the periphery can influence how individuals see each other and how they see themselves at work. In this paper, we present the looking-glass metaphor and use it to conceptualize AI knowledge systems as systems that reflect and distort, expanding our view on transparency requirements, implications and challenges. We formulate transparency as a key mediator in shaping different ways of seeing, including seeing into the system, which unveils its capabilities, limitations and behavior, and seeing through the system, which shapes workers' perceptions of their own contributions and others within the organization. Recognizing the sociotechnical nature of these systems, we identify three transparency dimensions necessary to realize the value of AI knowledge systems, namely system transparency, procedural transparency and transparency of outcomes. We discuss key challenges hindering the implementation of these forms of transparency, bringing to light the wider sociotechnical gap and highlighting directions for future Computer-supported Cooperative Work (CSCW) research.
1902.06950
Dominic Orchard
Li-yao Xia, Dominic Orchard, Meng Wang
Composing bidirectional programs monadically (with appendices)
Provides the appendices of the paper, which appears in the proceedings of European Symposium on Programming (ESOP) 2019
null
null
null
cs.PL
http://creativecommons.org/licenses/by/4.0/
Software frequently converts data from one representation to another and vice versa. Naively specifying both conversion directions separately is error prone and introduces conceptual duplication. Instead, bidirectional programming techniques allow programs to be written which can be interpreted in both directions. However, these techniques often employ unfamiliar programming idioms via restricted, specialised combinator libraries. Instead, we introduce a framework for composing bidirectional programs monadically, enabling bidirectional programming with familiar abstractions in functional languages such as Haskell. We demonstrate the generality of our approach applied to parsers/printers, lenses, and generators/predicates. We show how to leverage compositionality and equational reasoning for the verification of round-tripping properties for such monadic bidirectional programs.
[ { "created": "Tue, 19 Feb 2019 08:47:23 GMT", "version": "v1" } ]
2019-02-20
[ [ "Xia", "Li-yao", "" ], [ "Orchard", "Dominic", "" ], [ "Wang", "Meng", "" ] ]
Software frequently converts data from one representation to another and vice versa. Naively specifying both conversion directions separately is error prone and introduces conceptual duplication. Instead, bidirectional programming techniques allow programs to be written which can be interpreted in both directions. However, these techniques often employ unfamiliar programming idioms via restricted, specialised combinator libraries. Instead, we introduce a framework for composing bidirectional programs monadically, enabling bidirectional programming with familiar abstractions in functional languages such as Haskell. We demonstrate the generality of our approach applied to parsers/printers, lenses, and generators/predicates. We show how to leverage compositionality and equational reasoning for the verification of round-tripping properties for such monadic bidirectional programs.
1309.0752
Dr. Nadeem Javaid
N. Javaid, O. Rehman, N. Alrajeh, Z. A. Khan, B. Manzoor, S. Ahmed
AID: An Energy Efficient Decoding Scheme for LDPC Codes in Wireless Body Area Sensor Networks
2013 International Workshop on Communications and Sensor Networks (ComSense-2013), Niagara Falls, Ontario, Canada on October 21-24, 2013 in conjunction with 4th International Conference on Emerging Ubiquitous Systems and Pervasive Networks (EUSPN-2013)
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
One of the major challenges in Wireless Body Area Networks (WBANs) is to prolong the lifetime of network. Traditional research work focuses on minimizing transmit power, however, in the case of short range communication the consumption power in decoding is significantly larger than transmit power. This paper investigates the minimization of total power consumption by reducing the decoding power consumption. For achieving a desired Bit Error Rate (BER), we introduce some fundamental results on the basis of iterative message-passing algorithms for Low Density Parity Check Code (LDPC). To reduce energy dissipation in decoder, LDPC based coded communications between sensors are considered. Moreover, we evaluate the performance of LDPC at different code rates and introduce Adaptive Iterative Decoding (AID) by exploiting threshold on the number of iterations for a certain BER. In iterative LDPC decoding, the total energy consumption of network is reduced by 20 to 25 percent.
[ { "created": "Tue, 3 Sep 2013 17:30:48 GMT", "version": "v1" } ]
2013-09-04
[ [ "Javaid", "N.", "" ], [ "Rehman", "O.", "" ], [ "Alrajeh", "N.", "" ], [ "Khan", "Z. A.", "" ], [ "Manzoor", "B.", "" ], [ "Ahmed", "S.", "" ] ]
One of the major challenges in Wireless Body Area Networks (WBANs) is to prolong the lifetime of network. Traditional research work focuses on minimizing transmit power, however, in the case of short range communication the consumption power in decoding is significantly larger than transmit power. This paper investigates the minimization of total power consumption by reducing the decoding power consumption. For achieving a desired Bit Error Rate (BER), we introduce some fundamental results on the basis of iterative message-passing algorithms for Low Density Parity Check Code (LDPC). To reduce energy dissipation in decoder, LDPC based coded communications between sensors are considered. Moreover, we evaluate the performance of LDPC at different code rates and introduce Adaptive Iterative Decoding (AID) by exploiting threshold on the number of iterations for a certain BER. In iterative LDPC decoding, the total energy consumption of network is reduced by 20 to 25 percent.
2310.00535
Yuandong Tian
Yuandong Tian, Yiping Wang, Zhenyu Zhang, Beidi Chen, Simon Du
JoMA: Demystifying Multilayer Transformers via JOint Dynamics of MLP and Attention
ICLR'24 camera ready. Improve theorem 3 and theorem 4. Polish writing and add code link
null
null
null
cs.LG cs.AI cs.CL
http://creativecommons.org/licenses/by/4.0/
We propose Joint MLP/Attention (JoMA) dynamics, a novel mathematical framework to understand the training procedure of multilayer Transformer architectures. This is achieved by integrating out the self-attention layer in Transformers, producing a modified dynamics of MLP layers only. JoMA removes unrealistic assumptions in previous analysis (e.g., lack of residual connection) and predicts that the attention first becomes sparse (to learn salient tokens), then dense (to learn less salient tokens) in the presence of nonlinear activations, while in the linear case, it is consistent with existing works that show attention becomes sparse over time. We leverage JoMA to qualitatively explains how tokens are combined to form hierarchies in multilayer Transformers, when the input tokens are generated by a latent hierarchical generative model. Experiments on models trained from real-world dataset (Wikitext2/Wikitext103) and various pre-trained models (OPT, Pythia) verify our theoretical findings. Code can be found in https://github.com/facebookresearch/luckmatters/tree/yuandong3.
[ { "created": "Sun, 1 Oct 2023 01:21:35 GMT", "version": "v1" }, { "created": "Tue, 3 Oct 2023 04:23:26 GMT", "version": "v2" }, { "created": "Fri, 15 Mar 2024 02:03:21 GMT", "version": "v3" } ]
2024-03-18
[ [ "Tian", "Yuandong", "" ], [ "Wang", "Yiping", "" ], [ "Zhang", "Zhenyu", "" ], [ "Chen", "Beidi", "" ], [ "Du", "Simon", "" ] ]
We propose Joint MLP/Attention (JoMA) dynamics, a novel mathematical framework to understand the training procedure of multilayer Transformer architectures. This is achieved by integrating out the self-attention layer in Transformers, producing a modified dynamics of MLP layers only. JoMA removes unrealistic assumptions in previous analysis (e.g., lack of residual connection) and predicts that the attention first becomes sparse (to learn salient tokens), then dense (to learn less salient tokens) in the presence of nonlinear activations, while in the linear case, it is consistent with existing works that show attention becomes sparse over time. We leverage JoMA to qualitatively explains how tokens are combined to form hierarchies in multilayer Transformers, when the input tokens are generated by a latent hierarchical generative model. Experiments on models trained from real-world dataset (Wikitext2/Wikitext103) and various pre-trained models (OPT, Pythia) verify our theoretical findings. Code can be found in https://github.com/facebookresearch/luckmatters/tree/yuandong3.
2407.19402
Xihua Sheng
Xihua Sheng, Chuanbo Tang, Li Li, Dong Liu, Feng Wu
NVC-1B: A Large Neural Video Coding Model
null
null
null
null
cs.CV eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The emerging large models have achieved notable progress in the fields of natural language processing and computer vision. However, large models for neural video coding are still unexplored. In this paper, we try to explore how to build a large neural video coding model. Based on a small baseline model, we gradually scale up the model sizes of its different coding parts, including the motion encoder-decoder, motion entropy model, contextual encoder-decoder, contextual entropy model, and temporal context mining module, and analyze the influence of model sizes on video compression performance. Then, we explore to use different architectures, including CNN, mixed CNN-Transformer, and Transformer architectures, to implement the neural video coding model and analyze the influence of model architectures on video compression performance. Based on our exploration results, we design the first neural video coding model with more than 1 billion parameters -- NVC-1B. Experimental results show that our proposed large model achieves a significant video compression performance improvement over the small baseline model, and represents the state-of-the-art compression efficiency. We anticipate large models may bring up the video coding technologies to the next level.
[ { "created": "Sun, 28 Jul 2024 05:12:22 GMT", "version": "v1" } ]
2024-07-30
[ [ "Sheng", "Xihua", "" ], [ "Tang", "Chuanbo", "" ], [ "Li", "Li", "" ], [ "Liu", "Dong", "" ], [ "Wu", "Feng", "" ] ]
The emerging large models have achieved notable progress in the fields of natural language processing and computer vision. However, large models for neural video coding are still unexplored. In this paper, we try to explore how to build a large neural video coding model. Based on a small baseline model, we gradually scale up the model sizes of its different coding parts, including the motion encoder-decoder, motion entropy model, contextual encoder-decoder, contextual entropy model, and temporal context mining module, and analyze the influence of model sizes on video compression performance. Then, we explore to use different architectures, including CNN, mixed CNN-Transformer, and Transformer architectures, to implement the neural video coding model and analyze the influence of model architectures on video compression performance. Based on our exploration results, we design the first neural video coding model with more than 1 billion parameters -- NVC-1B. Experimental results show that our proposed large model achieves a significant video compression performance improvement over the small baseline model, and represents the state-of-the-art compression efficiency. We anticipate large models may bring up the video coding technologies to the next level.
2208.02504
Usman Akhtar Dr
Usman Akhtar, Rafal Kucharski
Exploring Computational Complexity Of Ride-Pooling Problems
13 pages, 7 figures, Submitted to The Transportation Research Board (TRB), Annual Meeting (102nd)
null
null
null
cs.DS cs.PF
http://creativecommons.org/licenses/by-nc-nd/4.0/
Ride-pooling is computationally challenging. The number of feasible rides grows with the number of travelers and the degree (capacity of the vehicle to perform a pooled ride) and quickly explodes to the sizes making the problem not solvable analytically. In practice, heuristics are applied to limit the number of searches, e.g., maximal detour and delay, or (like we use in this study) attractive rides (for which detour and delay are at least compensated with the discount). Nevertheless, the challenge to solve the ride-pooling remains strongly sensitive to the problem settings. Here, we explore it in more detail and provide an experimental underpinning to this open research problem. We trace how the size of the search space and computation time needed to solve the ride-pooling problem grows with the increasing demand and greater discounts offered for pooling. We run over 100 practical experiments in Amsterdam with 10-minute batches of trip requests up to 3600 trips per hour and trace how challenging it is to propose the solution to the pooling problem with our ExMAS algorithm. We observed strong, non-linear trends and identified the limits beyond which the problem exploded and our algorithm failed to compute. Notably, we found that the demand level (number of trip requests) is less critical than the discount. The search space grows exponentially and quickly reaches huge levels. However, beyond some level, the greater size of the ride-pooling problem does not translate into greater efficiency of pooling. Which opens the opportunity for further search space reductions.
[ { "created": "Thu, 4 Aug 2022 07:30:30 GMT", "version": "v1" } ]
2022-08-05
[ [ "Akhtar", "Usman", "" ], [ "Kucharski", "Rafal", "" ] ]
Ride-pooling is computationally challenging. The number of feasible rides grows with the number of travelers and the degree (capacity of the vehicle to perform a pooled ride) and quickly explodes to the sizes making the problem not solvable analytically. In practice, heuristics are applied to limit the number of searches, e.g., maximal detour and delay, or (like we use in this study) attractive rides (for which detour and delay are at least compensated with the discount). Nevertheless, the challenge to solve the ride-pooling remains strongly sensitive to the problem settings. Here, we explore it in more detail and provide an experimental underpinning to this open research problem. We trace how the size of the search space and computation time needed to solve the ride-pooling problem grows with the increasing demand and greater discounts offered for pooling. We run over 100 practical experiments in Amsterdam with 10-minute batches of trip requests up to 3600 trips per hour and trace how challenging it is to propose the solution to the pooling problem with our ExMAS algorithm. We observed strong, non-linear trends and identified the limits beyond which the problem exploded and our algorithm failed to compute. Notably, we found that the demand level (number of trip requests) is less critical than the discount. The search space grows exponentially and quickly reaches huge levels. However, beyond some level, the greater size of the ride-pooling problem does not translate into greater efficiency of pooling. Which opens the opportunity for further search space reductions.
2008.12735
Donald Honeycutt
Donald R. Honeycutt, Mahsan Nourani, Eric D. Ragan
Soliciting Human-in-the-Loop User Feedback for Interactive Machine Learning Reduces User Trust and Impressions of Model Accuracy
Accepted and to appear in the Proceedings of the AAAI Conference on Human Computation and Crowdsourcing (HCOMP) 2020
null
null
null
cs.HC cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Mixed-initiative systems allow users to interactively provide feedback to potentially improve system performance. Human feedback can correct model errors and update model parameters to dynamically adapt to changing data. Additionally, many users desire the ability to have a greater level of control and fix perceived flaws in systems they rely on. However, how the ability to provide feedback to autonomous systems influences user trust is a largely unexplored area of research. Our research investigates how the act of providing feedback can affect user understanding of an intelligent system and its accuracy. We present a controlled experiment using a simulated object detection system with image data to study the effects of interactive feedback collection on user impressions. The results show that providing human-in-the-loop feedback lowered both participants' trust in the system and their perception of system accuracy, regardless of whether the system accuracy improved in response to their feedback. These results highlight the importance of considering the effects of allowing end-user feedback on user trust when designing intelligent systems.
[ { "created": "Fri, 28 Aug 2020 16:46:41 GMT", "version": "v1" } ]
2020-08-31
[ [ "Honeycutt", "Donald R.", "" ], [ "Nourani", "Mahsan", "" ], [ "Ragan", "Eric D.", "" ] ]
Mixed-initiative systems allow users to interactively provide feedback to potentially improve system performance. Human feedback can correct model errors and update model parameters to dynamically adapt to changing data. Additionally, many users desire the ability to have a greater level of control and fix perceived flaws in systems they rely on. However, how the ability to provide feedback to autonomous systems influences user trust is a largely unexplored area of research. Our research investigates how the act of providing feedback can affect user understanding of an intelligent system and its accuracy. We present a controlled experiment using a simulated object detection system with image data to study the effects of interactive feedback collection on user impressions. The results show that providing human-in-the-loop feedback lowered both participants' trust in the system and their perception of system accuracy, regardless of whether the system accuracy improved in response to their feedback. These results highlight the importance of considering the effects of allowing end-user feedback on user trust when designing intelligent systems.
1802.00304
Malika Bendechache
Malika Bendechache and M-Tahar Kechadi
Distributed Clustering Algorithm for Spatial Data Mining
6 pages. arXiv admin note: text overlap with arXiv:1704.03421
Spatial Data Mining and Geographical Knowledge Services (ICSDM), 2015 2nd IEEE International Conference on, pages 60--65, 2015
10.1109/ICSDM.2015.7298026
null
cs.DB
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Distributed data mining techniques and mainly distributed clustering are widely used in the last decade because they deal with very large and heterogeneous datasets which cannot be gathered centrally. Current distributed clustering approaches are normally generating global models by aggregating local results that are obtained on each site. While this approach mines the datasets on their locations the aggregation phase is complex, which may produce incorrect and ambiguous global clusters and therefore incorrect knowledge. In this paper we propose a new clustering approach for very large spatial datasets that are heterogeneous and distributed. The approach is based on K-means Algorithm but it generates the number of global clusters dynamically. Moreover, this approach uses an elaborated aggregation phase. The aggregation phase is designed in such a way that the overall process is efficient in time and memory allocation. Preliminary results show that the proposed approach produces high quality results and scales up well. We also compared it to two popular clustering algorithms and show that this approach is much more efficient.
[ { "created": "Thu, 1 Feb 2018 14:41:33 GMT", "version": "v1" } ]
2018-02-02
[ [ "Bendechache", "Malika", "" ], [ "Kechadi", "M-Tahar", "" ] ]
Distributed data mining techniques and mainly distributed clustering are widely used in the last decade because they deal with very large and heterogeneous datasets which cannot be gathered centrally. Current distributed clustering approaches are normally generating global models by aggregating local results that are obtained on each site. While this approach mines the datasets on their locations the aggregation phase is complex, which may produce incorrect and ambiguous global clusters and therefore incorrect knowledge. In this paper we propose a new clustering approach for very large spatial datasets that are heterogeneous and distributed. The approach is based on K-means Algorithm but it generates the number of global clusters dynamically. Moreover, this approach uses an elaborated aggregation phase. The aggregation phase is designed in such a way that the overall process is efficient in time and memory allocation. Preliminary results show that the proposed approach produces high quality results and scales up well. We also compared it to two popular clustering algorithms and show that this approach is much more efficient.
2109.02682
Abdulaziz Alaboudi
Abdulaziz Alaboudi, Thomas D. LaToza
Edit-Run Behavior in Programming and Debugging
VL/HCC 2021
null
null
null
cs.SE cs.HC
http://creativecommons.org/licenses/by/4.0/
As developers program and debug, they continuously edit and run their code, a behavior known as edit-run cycles. While techniques such as live programming are intended to support this behavior, little is known about the characteristics of edit-run cycles themselves. To bridge this gap, we analyzed 28 hours of programming and debugging work from 11 professional developers which encompassed over three thousand development activities. We mapped activities to edit or run steps, constructing 581 debugging and 207 programming edit-run cycles. We found that edit-run cycles are frequent. Developers edit and run the program, on average, 7 times before fixing a defect and twice before introducing a defect. Developers waited longer before again running the program when programming than debugging, with a mean cycle length of 3 minutes for programming and 1 minute for debugging. Most cycles involved an edit to a single file after which a developer ran the program to observe the impact on the final output. Edit-run cycles which included activities beyond edit and run, such as navigating between files, consulting resources, or interacting with other IDE features, were much longer, with a mean length of 5 minutes, rather than 1.5 minutes. We conclude with a discussion of design recommendations for tools to enable more fluidity in edit-run cycles.
[ { "created": "Mon, 6 Sep 2021 18:06:01 GMT", "version": "v1" } ]
2021-09-08
[ [ "Alaboudi", "Abdulaziz", "" ], [ "LaToza", "Thomas D.", "" ] ]
As developers program and debug, they continuously edit and run their code, a behavior known as edit-run cycles. While techniques such as live programming are intended to support this behavior, little is known about the characteristics of edit-run cycles themselves. To bridge this gap, we analyzed 28 hours of programming and debugging work from 11 professional developers which encompassed over three thousand development activities. We mapped activities to edit or run steps, constructing 581 debugging and 207 programming edit-run cycles. We found that edit-run cycles are frequent. Developers edit and run the program, on average, 7 times before fixing a defect and twice before introducing a defect. Developers waited longer before again running the program when programming than debugging, with a mean cycle length of 3 minutes for programming and 1 minute for debugging. Most cycles involved an edit to a single file after which a developer ran the program to observe the impact on the final output. Edit-run cycles which included activities beyond edit and run, such as navigating between files, consulting resources, or interacting with other IDE features, were much longer, with a mean length of 5 minutes, rather than 1.5 minutes. We conclude with a discussion of design recommendations for tools to enable more fluidity in edit-run cycles.
1902.10030
Hussam Qassim Mr.
Hussein A. Al-Barazanchi, Hussam Qassim, David Feinzimer, and Abhishek Verma
Residual-CNDS for Grand Challenge Scene Dataset
null
null
null
null
cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Increasing depth of convolutional neural networks (CNNs) is a highly promising method of increasing the accuracy of the (CNNs). Increased CNN depth will also result in increased layer count (parameters), leading to a slow backpropagation convergence prone to overfitting. We trained our model (Residual-CNDS) to classify very large-scale scene datasets MIT Places 205, and MIT Places 365-Standard. The outcome result from the two datasets proved our proposed model (Residual-CNDS) effectively handled the slow convergence, overfitting, and degradation. CNNs that include deep supervision (CNDS) add supplementary branches to the deep convolutional neural network in specified layers by calculating vanishing, effectively addressing delayed convergence and overfitting. Nevertheless, (CNDS) does not resolve degradation; hence, we add residual learning to the (CNDS) in certain layers after studying the best place in which to add it. With this approach we overcome degradation in the very deep network. We have built two models (Residual-CNDS 8), and (Residual-CNDS 10). Moreover, we tested our models on two large-scale datasets, and we compared our results with other recently introduced cutting-edge networks in the domain of top-1 and top-5 classification accuracy. As a result, both of models have shown good improvement, which supports the assertion that the addition of residual connections enhances network CNDS accuracy without adding any computation complexity.
[ { "created": "Sun, 13 Jan 2019 23:00:11 GMT", "version": "v1" } ]
2019-02-27
[ [ "Al-Barazanchi", "Hussein A.", "" ], [ "Qassim", "Hussam", "" ], [ "Feinzimer", "David", "" ], [ "Verma", "Abhishek", "" ] ]
Increasing depth of convolutional neural networks (CNNs) is a highly promising method of increasing the accuracy of the (CNNs). Increased CNN depth will also result in increased layer count (parameters), leading to a slow backpropagation convergence prone to overfitting. We trained our model (Residual-CNDS) to classify very large-scale scene datasets MIT Places 205, and MIT Places 365-Standard. The outcome result from the two datasets proved our proposed model (Residual-CNDS) effectively handled the slow convergence, overfitting, and degradation. CNNs that include deep supervision (CNDS) add supplementary branches to the deep convolutional neural network in specified layers by calculating vanishing, effectively addressing delayed convergence and overfitting. Nevertheless, (CNDS) does not resolve degradation; hence, we add residual learning to the (CNDS) in certain layers after studying the best place in which to add it. With this approach we overcome degradation in the very deep network. We have built two models (Residual-CNDS 8), and (Residual-CNDS 10). Moreover, we tested our models on two large-scale datasets, and we compared our results with other recently introduced cutting-edge networks in the domain of top-1 and top-5 classification accuracy. As a result, both of models have shown good improvement, which supports the assertion that the addition of residual connections enhances network CNDS accuracy without adding any computation complexity.
2302.08624
Kevin Scaria
Kevin Scaria and Himanshu Gupta and Siddharth Goyal and Saurabh Arjun Sawant and Swaroop Mishra and Chitta Baral
InstructABSA: Instruction Learning for Aspect Based Sentiment Analysis
4 pages, 3 figures, 9 tables, 9 appendix pages
null
null
null
cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce InstructABSA, an instruction learning paradigm for Aspect-Based Sentiment Analysis (ABSA) subtasks. Our method introduces positive, negative, and neutral examples to each training sample, and instruction tune the model (Tk-Instruct) for ABSA subtasks, yielding significant performance improvements. Experimental results on the Sem Eval 2014, 15, and 16 datasets demonstrate that InstructABSA outperforms the previous state-of-the-art (SOTA) approaches on Term Extraction (ATE), Sentiment Classification(ATSC) and Sentiment Pair Extraction (ASPE) subtasks. In particular, InstructABSA outperforms the previous state-of-the-art (SOTA) on the Rest14 ATE subtask by 5.69% points, the Rest15 ATSC subtask by 9.59% points, and the Lapt14 AOPE subtask by 3.37% points, surpassing 7x larger models. We also get competitive results on AOOE, AOPE, and AOSTE subtasks indicating strong generalization ability to all subtasks. Exploring sample efficiency reveals that just 50% train data is required to get competitive results with other instruction tuning approaches. Lastly, we assess the quality of instructions and observe that InstructABSA's performance experiences a decline of ~10% when adding misleading examples.
[ { "created": "Thu, 16 Feb 2023 23:29:22 GMT", "version": "v1" }, { "created": "Tue, 21 Feb 2023 06:53:41 GMT", "version": "v2" }, { "created": "Wed, 5 Apr 2023 04:44:43 GMT", "version": "v3" }, { "created": "Thu, 20 Apr 2023 05:57:12 GMT", "version": "v4" }, { "created": "Thu, 25 May 2023 02:13:10 GMT", "version": "v5" }, { "created": "Mon, 13 Nov 2023 17:56:19 GMT", "version": "v6" } ]
2023-11-14
[ [ "Scaria", "Kevin", "" ], [ "Gupta", "Himanshu", "" ], [ "Goyal", "Siddharth", "" ], [ "Sawant", "Saurabh Arjun", "" ], [ "Mishra", "Swaroop", "" ], [ "Baral", "Chitta", "" ] ]
We introduce InstructABSA, an instruction learning paradigm for Aspect-Based Sentiment Analysis (ABSA) subtasks. Our method introduces positive, negative, and neutral examples to each training sample, and instruction tune the model (Tk-Instruct) for ABSA subtasks, yielding significant performance improvements. Experimental results on the Sem Eval 2014, 15, and 16 datasets demonstrate that InstructABSA outperforms the previous state-of-the-art (SOTA) approaches on Term Extraction (ATE), Sentiment Classification(ATSC) and Sentiment Pair Extraction (ASPE) subtasks. In particular, InstructABSA outperforms the previous state-of-the-art (SOTA) on the Rest14 ATE subtask by 5.69% points, the Rest15 ATSC subtask by 9.59% points, and the Lapt14 AOPE subtask by 3.37% points, surpassing 7x larger models. We also get competitive results on AOOE, AOPE, and AOSTE subtasks indicating strong generalization ability to all subtasks. Exploring sample efficiency reveals that just 50% train data is required to get competitive results with other instruction tuning approaches. Lastly, we assess the quality of instructions and observe that InstructABSA's performance experiences a decline of ~10% when adding misleading examples.
2105.11927
Chong Peng
Yang Liu, Qian Zhang, Yongyong Chen, Qiang Cheng and Chong Peng
Hyperspectral Image Denoising with Log-Based Robust PCA
null
null
null
null
cs.CV eess.IV
http://creativecommons.org/licenses/by/4.0/
It is a challenging task to remove heavy and mixed types of noise from Hyperspectral images (HSIs). In this paper, we propose a novel nonconvex approach to RPCA for HSI denoising, which adopts the log-determinant rank approximation and a novel $\ell_{2,\log}$ norm, to restrict the low-rank or column-wise sparse properties for the component matrices, respectively.For the $\ell_{2,\log}$-regularized shrinkage problem, we develop an efficient, closed-form solution, which is named $\ell_{2,\log}$-shrinkage operator, which can be generally used in other problems. Extensive experiments on both simulated and real HSIs demonstrate the effectiveness of the proposed method in denoising HSIs.
[ { "created": "Tue, 25 May 2021 13:32:01 GMT", "version": "v1" } ]
2021-05-26
[ [ "Liu", "Yang", "" ], [ "Zhang", "Qian", "" ], [ "Chen", "Yongyong", "" ], [ "Cheng", "Qiang", "" ], [ "Peng", "Chong", "" ] ]
It is a challenging task to remove heavy and mixed types of noise from Hyperspectral images (HSIs). In this paper, we propose a novel nonconvex approach to RPCA for HSI denoising, which adopts the log-determinant rank approximation and a novel $\ell_{2,\log}$ norm, to restrict the low-rank or column-wise sparse properties for the component matrices, respectively.For the $\ell_{2,\log}$-regularized shrinkage problem, we develop an efficient, closed-form solution, which is named $\ell_{2,\log}$-shrinkage operator, which can be generally used in other problems. Extensive experiments on both simulated and real HSIs demonstrate the effectiveness of the proposed method in denoising HSIs.
2205.08383
Matheus Schmitz
Matheus Schmitz, Rehan Ahmed, Jimi Cao
Bias and Fairness on Multimodal Emotion Detection Algorithms
null
null
10.13140/RG.2.2.14341.01769
null
cs.LG cs.AI cs.CL cs.CV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Numerous studies have shown that machine learning algorithms can latch onto protected attributes such as race and gender and generate predictions that systematically discriminate against one or more groups. To date the majority of bias and fairness research has been on unimodal models. In this work, we explore the biases that exist in emotion recognition systems in relationship to the modalities utilized, and study how multimodal approaches affect system bias and fairness. We consider audio, text, and video modalities, as well as all possible multimodal combinations of those, and find that text alone has the least bias, and accounts for the majority of the models' performances, raising doubts about the worthiness of multimodal emotion recognition systems when bias and fairness are desired alongside model performance.
[ { "created": "Wed, 11 May 2022 20:03:25 GMT", "version": "v1" } ]
2022-05-18
[ [ "Schmitz", "Matheus", "" ], [ "Ahmed", "Rehan", "" ], [ "Cao", "Jimi", "" ] ]
Numerous studies have shown that machine learning algorithms can latch onto protected attributes such as race and gender and generate predictions that systematically discriminate against one or more groups. To date the majority of bias and fairness research has been on unimodal models. In this work, we explore the biases that exist in emotion recognition systems in relationship to the modalities utilized, and study how multimodal approaches affect system bias and fairness. We consider audio, text, and video modalities, as well as all possible multimodal combinations of those, and find that text alone has the least bias, and accounts for the majority of the models' performances, raising doubts about the worthiness of multimodal emotion recognition systems when bias and fairness are desired alongside model performance.
2406.19106
Francesco Cambria
Francesco Cambria, Francesco Invernici, Anna Bernasconi and Stefano Ceri
MINE GRAPH RULE: A New Cypher-like Operator for Mining Association Rules on Property Graphs
null
null
null
null
cs.DB
http://creativecommons.org/licenses/by-sa/4.0/
Mining information from graph databases is becoming overly important. To approach this problem, current methods focus on identifying subgraphs with specific topologies; as of today, no work has been focused on expressing jointly the syntax and semantics of mining operations over rich property graphs. We define MINE GRAPH RULE, a new operator for mining association rules from graph databases, by extending classical approaches used in relational databases and exploited by recommending systems. We describe the syntax and semantics of the operator, which is based on measuring the support and confidence of each rule, and then we provide several examples of increasing complexity on top of a realistic example; our operator embeds Cypher for expressing the mining conditions. MINE GRAPH RULE is implemented on top of Neo4j, the most successful graph database system; it takes advantage of built-in optimizations of the Neo4j engine, as well as optimizations that are defined in the context of relational association rules. Our implementation is available as a portable Neo4j plugin. At the end of our paper, we show the execution performance in a variety of settings, by varying the operators, the size of the graph, the ratio between node types, the method for creating relationships, and maximum support and confidence.
[ { "created": "Thu, 27 Jun 2024 11:33:16 GMT", "version": "v1" } ]
2024-06-28
[ [ "Cambria", "Francesco", "" ], [ "Invernici", "Francesco", "" ], [ "Bernasconi", "Anna", "" ], [ "Ceri", "Stefano", "" ] ]
Mining information from graph databases is becoming overly important. To approach this problem, current methods focus on identifying subgraphs with specific topologies; as of today, no work has been focused on expressing jointly the syntax and semantics of mining operations over rich property graphs. We define MINE GRAPH RULE, a new operator for mining association rules from graph databases, by extending classical approaches used in relational databases and exploited by recommending systems. We describe the syntax and semantics of the operator, which is based on measuring the support and confidence of each rule, and then we provide several examples of increasing complexity on top of a realistic example; our operator embeds Cypher for expressing the mining conditions. MINE GRAPH RULE is implemented on top of Neo4j, the most successful graph database system; it takes advantage of built-in optimizations of the Neo4j engine, as well as optimizations that are defined in the context of relational association rules. Our implementation is available as a portable Neo4j plugin. At the end of our paper, we show the execution performance in a variety of settings, by varying the operators, the size of the graph, the ratio between node types, the method for creating relationships, and maximum support and confidence.
1505.02898
Bikramjit Singh
Bikramjit Singh, Konstantinos Koufos, Olav Tirkkonen
Coordination protocol for inter-operator spectrum sharing based on spectrum usage favors
Published in proceedings of 23rd edition of European Conference on Networks and Communications (EuCNC), Bologna, Jun. 2014
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Currently, mobile network operators are allocated spectrum bands on an exclusive basis. While this approach facilitates interference control, it may also result in low spectrum utilization efficiency. Inter-operator spectrum sharing is a potential method to enhance spectrum utilization. In order to realize it, a protocol to coordinate the actions of operators is needed. We propose a spectrum sharing protocol which is distributed in nature, it does not require operator-specific information exchange and it incurs minimal communication overhead between the operators. Operators are still free to decide whether they share spectrum or not as the protocol is based on the book keeping of spectrum usage favors, asked and received by the operators. We show that operators can enhance their QoS in comparison with traditional orthogonal spectrum allocation while also maintaining reciprocity i.e. no operator benefits over the other in the long run. We demonstrate the usability of the proposed protocol in an indoor deployment scenario with frequent network load variations as expected to have in small cell deployments.
[ { "created": "Tue, 12 May 2015 08:10:44 GMT", "version": "v1" } ]
2015-05-13
[ [ "Singh", "Bikramjit", "" ], [ "Koufos", "Konstantinos", "" ], [ "Tirkkonen", "Olav", "" ] ]
Currently, mobile network operators are allocated spectrum bands on an exclusive basis. While this approach facilitates interference control, it may also result in low spectrum utilization efficiency. Inter-operator spectrum sharing is a potential method to enhance spectrum utilization. In order to realize it, a protocol to coordinate the actions of operators is needed. We propose a spectrum sharing protocol which is distributed in nature, it does not require operator-specific information exchange and it incurs minimal communication overhead between the operators. Operators are still free to decide whether they share spectrum or not as the protocol is based on the book keeping of spectrum usage favors, asked and received by the operators. We show that operators can enhance their QoS in comparison with traditional orthogonal spectrum allocation while also maintaining reciprocity i.e. no operator benefits over the other in the long run. We demonstrate the usability of the proposed protocol in an indoor deployment scenario with frequent network load variations as expected to have in small cell deployments.
1908.07820
XiaoKang Liu
Jianquan Li, Xiaokang Liu, Wenpeng Yin, Min Yang, Liqun Ma, Yaohong Jin
Empirical Evaluation of Multi-task Learning in Deep Neural Networks for Natural Language Processing
null
null
null
null
cs.CL cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Multi-Task Learning (MTL) aims at boosting the overall performance of each individual task by leveraging useful information contained in multiple related tasks. It has shown great success in natural language processing (NLP). Currently, a number of MLT architectures and learning mechanisms have been proposed for various NLP tasks. However, there is no systematic exploration and comparison of different MLT architectures and learning mechanisms for their strong performance in-depth. In this paper, we conduct a thorough examination of typical MTL methods on a broad range of representative NLP tasks. Our primary goal is to understand the merits and demerits of existing MTL methods in NLP tasks, thus devising new hybrid architectures intended to combine their strengths.
[ { "created": "Fri, 16 Aug 2019 03:16:40 GMT", "version": "v1" }, { "created": "Fri, 7 Aug 2020 08:06:18 GMT", "version": "v2" } ]
2020-08-10
[ [ "Li", "Jianquan", "" ], [ "Liu", "Xiaokang", "" ], [ "Yin", "Wenpeng", "" ], [ "Yang", "Min", "" ], [ "Ma", "Liqun", "" ], [ "Jin", "Yaohong", "" ] ]
Multi-Task Learning (MTL) aims at boosting the overall performance of each individual task by leveraging useful information contained in multiple related tasks. It has shown great success in natural language processing (NLP). Currently, a number of MLT architectures and learning mechanisms have been proposed for various NLP tasks. However, there is no systematic exploration and comparison of different MLT architectures and learning mechanisms for their strong performance in-depth. In this paper, we conduct a thorough examination of typical MTL methods on a broad range of representative NLP tasks. Our primary goal is to understand the merits and demerits of existing MTL methods in NLP tasks, thus devising new hybrid architectures intended to combine their strengths.
1704.01893
Lukas Holzbaur
Lukas Holzbaur, Hannes Bartz, Antonia Wachter-Zeh
Improved Decoding and Error Floor Analysis of Staircase Codes
null
null
10.1007/s10623-018-0587-x
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Staircase codes play an important role as error-correcting codes in optical communications. In this paper, a low-complexity method for resolving stall patterns when decoding staircase codes is described. Stall patterns are the dominating contributor to the error floor in the original decoding method. Our improvement is based on locating stall patterns by intersecting non-zero syndromes and flipping the corresponding bits. The approach effectively lowers the error floor and allows for a new range of block sizes to be considered for optical communications at a certain rate or, alternatively, a significantly decreased error floor for the same block size. Further, an improved error floor analysis is introduced which provides a more accurate estimation of the contributions to the error floor.
[ { "created": "Thu, 6 Apr 2017 15:39:52 GMT", "version": "v1" }, { "created": "Fri, 5 Jan 2018 08:36:43 GMT", "version": "v2" }, { "created": "Thu, 5 Jul 2018 13:49:30 GMT", "version": "v3" }, { "created": "Mon, 3 Dec 2018 09:26:43 GMT", "version": "v4" } ]
2018-12-04
[ [ "Holzbaur", "Lukas", "" ], [ "Bartz", "Hannes", "" ], [ "Wachter-Zeh", "Antonia", "" ] ]
Staircase codes play an important role as error-correcting codes in optical communications. In this paper, a low-complexity method for resolving stall patterns when decoding staircase codes is described. Stall patterns are the dominating contributor to the error floor in the original decoding method. Our improvement is based on locating stall patterns by intersecting non-zero syndromes and flipping the corresponding bits. The approach effectively lowers the error floor and allows for a new range of block sizes to be considered for optical communications at a certain rate or, alternatively, a significantly decreased error floor for the same block size. Further, an improved error floor analysis is introduced which provides a more accurate estimation of the contributions to the error floor.
1506.03340
Karl Moritz Hermann
Karl Moritz Hermann, Tom\'a\v{s} Ko\v{c}isk\'y, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman and Phil Blunsom
Teaching Machines to Read and Comprehend
Appears in: Advances in Neural Information Processing Systems 28 (NIPS 2015). 14 pages, 13 figures
null
null
null
cs.CL cs.AI cs.NE
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure.
[ { "created": "Wed, 10 Jun 2015 14:54:39 GMT", "version": "v1" }, { "created": "Thu, 1 Oct 2015 15:04:49 GMT", "version": "v2" }, { "created": "Thu, 19 Nov 2015 15:43:23 GMT", "version": "v3" } ]
2015-11-20
[ [ "Hermann", "Karl Moritz", "" ], [ "Kočiský", "Tomáš", "" ], [ "Grefenstette", "Edward", "" ], [ "Espeholt", "Lasse", "" ], [ "Kay", "Will", "" ], [ "Suleyman", "Mustafa", "" ], [ "Blunsom", "Phil", "" ] ]
Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure.
1911.12396
Marcelo Firer
Marcelo Firer
Alternative Metrics
This is a chapter for "Concise Encyclopedia of Coding Theory" to be published by CRC Press
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The main scope of this chapter is metrics defined for coding and decoding purposes, mainly for block codes.
[ { "created": "Wed, 27 Nov 2019 19:46:55 GMT", "version": "v1" } ]
2019-12-02
[ [ "Firer", "Marcelo", "" ] ]
The main scope of this chapter is metrics defined for coding and decoding purposes, mainly for block codes.
1701.05924
Maria Cabrera
Maria Cabrera, Richard Voyles, Juan Wachs
Coherency in One-Shot Gesture Recognition
This paper was submitted to a IEEE conference
null
null
null
cs.HC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
User's intentions may be expressed through spontaneous gesturing, which have been seen only a few times or never before. Recognizing such gestures involves one shot gesture learning. While most research has focused on the recognition of the gestures itself, recently new approaches were proposed to deal with gesture perception and production as part of the same problem. The framework presented in this work focuses on learning the process that leads to gesture generation, rather than mining the gesture's associated features. This is achieved using kinematic, cognitive and biomechanic characteristics of human interaction. These factors enable the artificial production of realistic gesture samples originated from a single observation. The generated samples are then used as training sets for different state-of-the-art classifiers. Performance is obtained first, by observing the machines' gesture recognition percentages. Then, performance is computed by the human recognition from gestures performed by robots. Based on these two scenarios, a composite new metric of coherency is proposed relating to the amount of agreement between these two conditions. Experimental results provide an average recognition performance of 89.2% for the trained classifiers and 92.5% for the participants. Coherency in recognition was determined at 93.6%. While this new metric is not directly comparable to raw accuracy or other pure performance-based standard metrics, it provides a quantifier for validating how realistic the machine generated samples are and how accurate the resulting mimicry is.
[ { "created": "Fri, 20 Jan 2017 20:54:10 GMT", "version": "v1" } ]
2017-01-24
[ [ "Cabrera", "Maria", "" ], [ "Voyles", "Richard", "" ], [ "Wachs", "Juan", "" ] ]
User's intentions may be expressed through spontaneous gesturing, which have been seen only a few times or never before. Recognizing such gestures involves one shot gesture learning. While most research has focused on the recognition of the gestures itself, recently new approaches were proposed to deal with gesture perception and production as part of the same problem. The framework presented in this work focuses on learning the process that leads to gesture generation, rather than mining the gesture's associated features. This is achieved using kinematic, cognitive and biomechanic characteristics of human interaction. These factors enable the artificial production of realistic gesture samples originated from a single observation. The generated samples are then used as training sets for different state-of-the-art classifiers. Performance is obtained first, by observing the machines' gesture recognition percentages. Then, performance is computed by the human recognition from gestures performed by robots. Based on these two scenarios, a composite new metric of coherency is proposed relating to the amount of agreement between these two conditions. Experimental results provide an average recognition performance of 89.2% for the trained classifiers and 92.5% for the participants. Coherency in recognition was determined at 93.6%. While this new metric is not directly comparable to raw accuracy or other pure performance-based standard metrics, it provides a quantifier for validating how realistic the machine generated samples are and how accurate the resulting mimicry is.
2206.04928
Mohit Vaishnav
Mohit Vaishnav, Thomas Serre
GAMR: A Guided Attention Model for (visual) Reasoning
null
Eleventh International Conference on Learning Representations (ICLR) 2023
null
null
cs.AI cs.LG cs.NE cs.SC
http://creativecommons.org/licenses/by-nc-sa/4.0/
Humans continue to outperform modern AI systems in their ability to flexibly parse and understand complex visual scenes. Here, we present a novel module for visual reasoning, the Guided Attention Model for (visual) Reasoning (GAMR), which instantiates an active vision theory -- positing that the brain solves complex visual reasoning problems dynamically -- via sequences of attention shifts to select and route task-relevant visual information into memory. Experiments on an array of visual reasoning tasks and datasets demonstrate GAMR's ability to learn visual routines in a robust and sample-efficient manner. In addition, GAMR is shown to be capable of zero-shot generalization on completely novel reasoning tasks. Overall, our work provides computational support for cognitive theories that postulate the need for a critical interplay between attention and memory to dynamically maintain and manipulate task-relevant visual information to solve complex visual reasoning tasks.
[ { "created": "Fri, 10 Jun 2022 07:52:06 GMT", "version": "v1" }, { "created": "Mon, 13 Jun 2022 17:52:57 GMT", "version": "v2" }, { "created": "Wed, 21 Sep 2022 10:17:20 GMT", "version": "v3" }, { "created": "Thu, 22 Sep 2022 11:57:12 GMT", "version": "v4" }, { "created": "Tue, 21 Mar 2023 15:35:50 GMT", "version": "v5" } ]
2023-03-22
[ [ "Vaishnav", "Mohit", "" ], [ "Serre", "Thomas", "" ] ]
Humans continue to outperform modern AI systems in their ability to flexibly parse and understand complex visual scenes. Here, we present a novel module for visual reasoning, the Guided Attention Model for (visual) Reasoning (GAMR), which instantiates an active vision theory -- positing that the brain solves complex visual reasoning problems dynamically -- via sequences of attention shifts to select and route task-relevant visual information into memory. Experiments on an array of visual reasoning tasks and datasets demonstrate GAMR's ability to learn visual routines in a robust and sample-efficient manner. In addition, GAMR is shown to be capable of zero-shot generalization on completely novel reasoning tasks. Overall, our work provides computational support for cognitive theories that postulate the need for a critical interplay between attention and memory to dynamically maintain and manipulate task-relevant visual information to solve complex visual reasoning tasks.
2303.05862
Margherita Bert\`e
Margherita Bert\`e, Kyriaki Kalimeri, Daniela Paolotti
Monitoring Gender Gaps via LinkedIn Advertising Estimates: the case study of Italy
10 pages
In Proceedings of the ACM Web Science Conference 2023 (WebSci '23), April 30-May 1, 2023, Evanston, TX, USA. ACM, New York, NY, USA, 10 pages
10.1145/3578503.3583629
null
cs.CY
http://creativecommons.org/licenses/by/4.0/
Women remain underrepresented in the labour market. Although significant advancements are being made to increase female participation in the workforce, the gender gap is still far from being bridged. We contribute to the growing literature on gender inequalities in the labour market, evaluating the potential of the LinkedIn estimates to monitor the evolution of the gender gaps sustainably, complementing the official data sources. In particular, assessing the labour market patterns at a subnational level in Italy. Our findings show that the LinkedIn estimates accurately capture the gender disparities in Italy regarding sociodemographic attributes such as gender, age, geographic location, seniority, and industry category. At the same time, we assess data biases such as the digitalisation gap, which impacts the representativity of the workforce in an imbalanced manner, confirming that women are under-represented in Southern Italy. Additionally to confirming the gender disparities to the official census, LinkedIn estimates are a valuable tool to provide dynamic insights; we showed an immigration flow of highly skilled women, predominantly from the South. Digital surveillance of gender inequalities with detailed and timely data is particularly significant to enable policymakers to tailor impactful campaigns.
[ { "created": "Fri, 10 Mar 2023 11:32:45 GMT", "version": "v1" } ]
2023-03-13
[ [ "Bertè", "Margherita", "" ], [ "Kalimeri", "Kyriaki", "" ], [ "Paolotti", "Daniela", "" ] ]
Women remain underrepresented in the labour market. Although significant advancements are being made to increase female participation in the workforce, the gender gap is still far from being bridged. We contribute to the growing literature on gender inequalities in the labour market, evaluating the potential of the LinkedIn estimates to monitor the evolution of the gender gaps sustainably, complementing the official data sources. In particular, assessing the labour market patterns at a subnational level in Italy. Our findings show that the LinkedIn estimates accurately capture the gender disparities in Italy regarding sociodemographic attributes such as gender, age, geographic location, seniority, and industry category. At the same time, we assess data biases such as the digitalisation gap, which impacts the representativity of the workforce in an imbalanced manner, confirming that women are under-represented in Southern Italy. Additionally to confirming the gender disparities to the official census, LinkedIn estimates are a valuable tool to provide dynamic insights; we showed an immigration flow of highly skilled women, predominantly from the South. Digital surveillance of gender inequalities with detailed and timely data is particularly significant to enable policymakers to tailor impactful campaigns.
2104.00236
Jack Li
Jianhua Li, Ximeng Liu, Jiong Jin, Shui Yu
Too Expensive to Attack: A Joint Defense Framework to Mitigate Distributed Attacks for the Internet of Things Grid
10 pages, 10 figures, 5 tables
null
null
null
cs.CR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The distributed denial of service (DDoS) attack is detrimental to businesses and individuals as we are heavily relying on the Internet. Due to remarkable profits, crackers favor DDoS as cybersecurity weapons in attacking servers, computers, IoT devices, and even the entire Internet. Many current detection and mitigation solutions concentrate on specific technologies in combating DDoS, whereas the attacking expense and the cross-defender collaboration have not drawn enough attention. Under this circumstance, we revisit the DDoS attack and defense in terms of attacking cost and populations of both parties, proposing a joint defense framework to incur higher attacking expense in a grid of Internet service providers (ISPs), businesses, individuals, and third-party organizations (IoT Grid). Meanwhile, the defender's cost does not grow much during combats. The skyrocket of attacking expense discourages profit-driven attackers from launching further attacks effectively. The quantitative evaluation and experimental assessment reinforce the effectiveness of our framework.
[ { "created": "Thu, 1 Apr 2021 03:40:29 GMT", "version": "v1" } ]
2021-04-02
[ [ "Li", "Jianhua", "" ], [ "Liu", "Ximeng", "" ], [ "Jin", "Jiong", "" ], [ "Yu", "Shui", "" ] ]
The distributed denial of service (DDoS) attack is detrimental to businesses and individuals as we are heavily relying on the Internet. Due to remarkable profits, crackers favor DDoS as cybersecurity weapons in attacking servers, computers, IoT devices, and even the entire Internet. Many current detection and mitigation solutions concentrate on specific technologies in combating DDoS, whereas the attacking expense and the cross-defender collaboration have not drawn enough attention. Under this circumstance, we revisit the DDoS attack and defense in terms of attacking cost and populations of both parties, proposing a joint defense framework to incur higher attacking expense in a grid of Internet service providers (ISPs), businesses, individuals, and third-party organizations (IoT Grid). Meanwhile, the defender's cost does not grow much during combats. The skyrocket of attacking expense discourages profit-driven attackers from launching further attacks effectively. The quantitative evaluation and experimental assessment reinforce the effectiveness of our framework.
1701.00146
Erik Demaine
Jeffrey Bosboom, Erik D. Demaine, Martin L. Demaine, Adam Hesterberg, Pasin Manurangsi, Anak Yodpinyanee
Even $1 \times n$ Edge-Matching and Jigsaw Puzzles are Really Hard
22 pages, 9 figures
null
null
null
cs.CC cs.CG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We prove the computational intractability of rotating and placing $n$ square tiles into a $1 \times n$ array such that adjacent tiles are compatible--either equal edge colors, as in edge-matching puzzles, or matching tab/pocket shapes, as in jigsaw puzzles. Beyond basic NP-hardness, we prove that it is NP-hard even to approximately maximize the number of placed tiles (allowing blanks), while satisfying the compatibility constraint between nonblank tiles, within a factor of 0.9999999851. (On the other hand, there is an easy $1 \over 2$-approximation.) This is the first (correct) proof of inapproximability for edge-matching and jigsaw puzzles. Along the way, we prove NP-hardness of distinguishing, for a directed graph on $n$ nodes, between having a Hamiltonian path (length $n-1$) and having at most $0.999999284 (n-1)$ edges that form a vertex-disjoint union of paths. We use this gap hardness and gap-preserving reductions to establish similar gap hardness for $1 \times n$ jigsaw and edge-matching puzzles.
[ { "created": "Sat, 31 Dec 2016 17:05:53 GMT", "version": "v1" } ]
2017-01-03
[ [ "Bosboom", "Jeffrey", "" ], [ "Demaine", "Erik D.", "" ], [ "Demaine", "Martin L.", "" ], [ "Hesterberg", "Adam", "" ], [ "Manurangsi", "Pasin", "" ], [ "Yodpinyanee", "Anak", "" ] ]
We prove the computational intractability of rotating and placing $n$ square tiles into a $1 \times n$ array such that adjacent tiles are compatible--either equal edge colors, as in edge-matching puzzles, or matching tab/pocket shapes, as in jigsaw puzzles. Beyond basic NP-hardness, we prove that it is NP-hard even to approximately maximize the number of placed tiles (allowing blanks), while satisfying the compatibility constraint between nonblank tiles, within a factor of 0.9999999851. (On the other hand, there is an easy $1 \over 2$-approximation.) This is the first (correct) proof of inapproximability for edge-matching and jigsaw puzzles. Along the way, we prove NP-hardness of distinguishing, for a directed graph on $n$ nodes, between having a Hamiltonian path (length $n-1$) and having at most $0.999999284 (n-1)$ edges that form a vertex-disjoint union of paths. We use this gap hardness and gap-preserving reductions to establish similar gap hardness for $1 \times n$ jigsaw and edge-matching puzzles.
1002.1691
Rdv Ijcsis
Sumon Kumar Debnath, Foez Ahmed, Nayeema Islam
Performance Evaluation of Unicast and Broadcast Mobile Ad hoc Network Routing Protocols
7 Pages IEEE format, International Journal of Computer Science and Information Security, IJCSIS January 2010, ISSN 1947 5500, http://sites.google.com/site/ijcsis/
International Journal of Computer Science and Information Security, IJCSIS, Vol. 7, No. 1, pp. 40-46, January 2010, USA
null
Computer Science Volume 7 ISSN 19475500
cs.NI cs.PF
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Efficient routing mechanism is a challenging issue for group oriented computing in Mobile Ad Hoc Networks (MANETs). The ability of MANETs to support adequate Quality of Service (QoS) for group communication is limited by the ability of the underlying ad-hoc routing protocols to provide consistent behavior despite the dynamic properties of mobile computing devices. In MANET QoS requirements can be quantified in terms of Packet Delivery Ratio (PDR), Data Latency, Packet Loss Probability, Routing Overhead, Medium Access Control (MAC) Overhead and Data Throughput etc. This paper presents an in depth study of one to many and many to many communications in MANETs and provides a comparative performance evaluation of unicast and broadcast routing protocols. Dynamic Source Routing protocol (DSR) is used as unicast protocol and BCAST is used to represent broadcast protocol. The performance differentials are analyzed using ns2 network simulator varying multicast group size (number of data senders and data receivers). Both protocols are simulated with identical traffic loads and mobility models. Simulation result shows that BCAST performs better than DSR in most cases.
[ { "created": "Mon, 8 Feb 2010 19:12:53 GMT", "version": "v1" } ]
2010-02-09
[ [ "Debnath", "Sumon Kumar", "" ], [ "Ahmed", "Foez", "" ], [ "Islam", "Nayeema", "" ] ]
Efficient routing mechanism is a challenging issue for group oriented computing in Mobile Ad Hoc Networks (MANETs). The ability of MANETs to support adequate Quality of Service (QoS) for group communication is limited by the ability of the underlying ad-hoc routing protocols to provide consistent behavior despite the dynamic properties of mobile computing devices. In MANET QoS requirements can be quantified in terms of Packet Delivery Ratio (PDR), Data Latency, Packet Loss Probability, Routing Overhead, Medium Access Control (MAC) Overhead and Data Throughput etc. This paper presents an in depth study of one to many and many to many communications in MANETs and provides a comparative performance evaluation of unicast and broadcast routing protocols. Dynamic Source Routing protocol (DSR) is used as unicast protocol and BCAST is used to represent broadcast protocol. The performance differentials are analyzed using ns2 network simulator varying multicast group size (number of data senders and data receivers). Both protocols are simulated with identical traffic loads and mobility models. Simulation result shows that BCAST performs better than DSR in most cases.
1903.01298
Elvin Isufi
Elvin Isufi, Fernando Gama, Alejandro Ribeiro
Generalizing Graph Convolutional Neural Networks with Edge-Variant Recursions on Graphs
submitted to EUSIPCO 2019
null
null
null
cs.LG eess.SP stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper reviews graph convolutional neural networks (GCNNs) through the lens of edge-variant graph filters. The edge-variant graph filter is a finite order, linear, and local recursion that allows each node, in each iteration, to weigh differently the information of its neighbors. By exploiting this recursion, we formulate a general framework for GCNNs which considers state-of-the-art solutions as particular cases. This framework results useful to i) understand the tradeoff between local detail and the number of parameters of each solution and ii) provide guidelines for developing a myriad of novel approaches that can be implemented locally in the vertex domain. One of such approaches is presented here showing superior performance w.r.t. current alternatives in graph signal classification problems.
[ { "created": "Mon, 4 Mar 2019 15:05:36 GMT", "version": "v1" } ]
2019-03-05
[ [ "Isufi", "Elvin", "" ], [ "Gama", "Fernando", "" ], [ "Ribeiro", "Alejandro", "" ] ]
This paper reviews graph convolutional neural networks (GCNNs) through the lens of edge-variant graph filters. The edge-variant graph filter is a finite order, linear, and local recursion that allows each node, in each iteration, to weigh differently the information of its neighbors. By exploiting this recursion, we formulate a general framework for GCNNs which considers state-of-the-art solutions as particular cases. This framework results useful to i) understand the tradeoff between local detail and the number of parameters of each solution and ii) provide guidelines for developing a myriad of novel approaches that can be implemented locally in the vertex domain. One of such approaches is presented here showing superior performance w.r.t. current alternatives in graph signal classification problems.
1809.00952
Irvin Aloise
Irvin Aloise and Giorgio Grisetti
Matrix Difference in Pose-Graph Optimization
10 pages, 7 figures, source: https://srrg.gitlab.io/g2o_chordal_plugin.html
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Pose-Graph optimization is a crucial component of many modern SLAM systems. Most prominent state of the art systems address this problem by iterative non-linear least squares. Both number of iterations and convergence basin of these approaches depend on the error functions used to describe the problem. The smoother and more convex the error function with respect to perturbations of the state variables, the better the least-squares solver will perform. In this paper we propose an alternative error function obtained by removing some non-linearities from the standard used one - i.e. the geodesic error function. Comparative experiments conducted on common benchmarking datasets confirm that our function is more robust to noise that affects the rotational component of the pose measurements and, thus, exhibits a larger convergence basin than the geodesic. Furthermore, its implementation is relatively easy compared to the geodesic distance. This property leads to rather simple derivatives and nice numerical properties of the Jacobians resulting from the effective computation of the quadratic approximation used by Gauss-Newton algorithm.
[ { "created": "Tue, 4 Sep 2018 13:41:01 GMT", "version": "v1" } ]
2018-09-05
[ [ "Aloise", "Irvin", "" ], [ "Grisetti", "Giorgio", "" ] ]
Pose-Graph optimization is a crucial component of many modern SLAM systems. Most prominent state of the art systems address this problem by iterative non-linear least squares. Both number of iterations and convergence basin of these approaches depend on the error functions used to describe the problem. The smoother and more convex the error function with respect to perturbations of the state variables, the better the least-squares solver will perform. In this paper we propose an alternative error function obtained by removing some non-linearities from the standard used one - i.e. the geodesic error function. Comparative experiments conducted on common benchmarking datasets confirm that our function is more robust to noise that affects the rotational component of the pose measurements and, thus, exhibits a larger convergence basin than the geodesic. Furthermore, its implementation is relatively easy compared to the geodesic distance. This property leads to rather simple derivatives and nice numerical properties of the Jacobians resulting from the effective computation of the quadratic approximation used by Gauss-Newton algorithm.
2302.05959
Yanheng Li
Yanheng Li, Lin Luoying, Xinyan Li, Yaxuan Mao, Ray Lc
"Nice to meet you!": Expressing Emotions with Movement Gestures and Textual Content in Automatic Handwriting Robots
HRI 2023 LBR
null
10.1145/3568294.3580045
null
cs.HC cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Text-writing robots have been used in assistive writing and drawing applications. However, robots do not convey emotional tones in the writing process due to the lack of behaviors humans typically adopt. To examine how people interpret designed robotic expressions of emotion through both movements and textual output, we used a pen-plotting robot to generate texts by performing human-like behaviors like stop-and-go, speed, and pressure variation. We examined how people convey emotion in the writing process by observing how they wrote in different emotional contexts. We then mapped these human expressions during writing to the handwriting robot and measured how well other participants understood the robot's affective expression. We found that textual output was the strongest determinant of participants' ability to perceive the robot's emotions, whereas parameters of gestural movements of the robots like speed, fluency, pressure, size, and acceleration could be useful for understanding the context of the writing expression.
[ { "created": "Sun, 12 Feb 2023 17:13:25 GMT", "version": "v1" } ]
2023-02-14
[ [ "Li", "Yanheng", "" ], [ "Luoying", "Lin", "" ], [ "Li", "Xinyan", "" ], [ "Mao", "Yaxuan", "" ], [ "Lc", "Ray", "" ] ]
Text-writing robots have been used in assistive writing and drawing applications. However, robots do not convey emotional tones in the writing process due to the lack of behaviors humans typically adopt. To examine how people interpret designed robotic expressions of emotion through both movements and textual output, we used a pen-plotting robot to generate texts by performing human-like behaviors like stop-and-go, speed, and pressure variation. We examined how people convey emotion in the writing process by observing how they wrote in different emotional contexts. We then mapped these human expressions during writing to the handwriting robot and measured how well other participants understood the robot's affective expression. We found that textual output was the strongest determinant of participants' ability to perceive the robot's emotions, whereas parameters of gestural movements of the robots like speed, fluency, pressure, size, and acceleration could be useful for understanding the context of the writing expression.
2107.07116
Feng Shi
Feng Shi, Chonghan Lee, Mohammad Khairul Bashar, Nikhil Shukla, Song-Chun Zhu and Vijaykrishnan Narayanan
Transformer-based Machine Learning for Fast SAT Solvers and Logic Synthesis
null
null
null
null
cs.NE cs.AI cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
CNF-based SAT and MaxSAT solvers are central to logic synthesis and verification systems. The increasing popularity of these constraint problems in electronic design automation encourages studies on different SAT problems and their properties for further computational efficiency. There has been both theoretical and practical success of modern Conflict-driven clause learning SAT solvers, which allows solving very large industrial instances in a relatively short amount of time. Recently, machine learning approaches provide a new dimension to solving this challenging problem. Neural symbolic models could serve as generic solvers that can be specialized for specific domains based on data without any changes to the structure of the model. In this work, we propose a one-shot model derived from the Transformer architecture to solve the MaxSAT problem, which is the optimization version of SAT where the goal is to satisfy the maximum number of clauses. Our model has a scale-free structure which could process varying size of instances. We use meta-path and self-attention mechanism to capture interactions among homogeneous nodes. We adopt cross-attention mechanisms on the bipartite graph to capture interactions among heterogeneous nodes. We further apply an iterative algorithm to our model to satisfy additional clauses, enabling a solution approaching that of an exact-SAT problem. The attention mechanisms leverage the parallelism for speedup. Our evaluation indicates improved speedup compared to heuristic approaches and improved completion rate compared to machine learning approaches.
[ { "created": "Thu, 15 Jul 2021 04:47:35 GMT", "version": "v1" } ]
2021-07-16
[ [ "Shi", "Feng", "" ], [ "Lee", "Chonghan", "" ], [ "Bashar", "Mohammad Khairul", "" ], [ "Shukla", "Nikhil", "" ], [ "Zhu", "Song-Chun", "" ], [ "Narayanan", "Vijaykrishnan", "" ] ]
CNF-based SAT and MaxSAT solvers are central to logic synthesis and verification systems. The increasing popularity of these constraint problems in electronic design automation encourages studies on different SAT problems and their properties for further computational efficiency. There has been both theoretical and practical success of modern Conflict-driven clause learning SAT solvers, which allows solving very large industrial instances in a relatively short amount of time. Recently, machine learning approaches provide a new dimension to solving this challenging problem. Neural symbolic models could serve as generic solvers that can be specialized for specific domains based on data without any changes to the structure of the model. In this work, we propose a one-shot model derived from the Transformer architecture to solve the MaxSAT problem, which is the optimization version of SAT where the goal is to satisfy the maximum number of clauses. Our model has a scale-free structure which could process varying size of instances. We use meta-path and self-attention mechanism to capture interactions among homogeneous nodes. We adopt cross-attention mechanisms on the bipartite graph to capture interactions among heterogeneous nodes. We further apply an iterative algorithm to our model to satisfy additional clauses, enabling a solution approaching that of an exact-SAT problem. The attention mechanisms leverage the parallelism for speedup. Our evaluation indicates improved speedup compared to heuristic approaches and improved completion rate compared to machine learning approaches.
2106.13679
Giovanni Trappolini
Giovanni Trappolini, Luca Cosmo, Luca Moschella, Riccardo Marin, Simone Melzi, Emanuele Rodol\`a
Shape registration in the time of transformers
null
null
null
null
cs.CV cs.GR cs.LG
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper, we propose a transformer-based procedure for the efficient registration of non-rigid 3D point clouds. The proposed approach is data-driven and adopts for the first time the transformer architecture in the registration task. Our method is general and applies to different settings. Given a fixed template with some desired properties (e.g. skinning weights or other animation cues), we can register raw acquired data to it, thereby transferring all the template properties to the input geometry. Alternatively, given a pair of shapes, our method can register the first onto the second (or vice-versa), obtaining a high-quality dense correspondence between the two. In both contexts, the quality of our results enables us to target real applications such as texture transfer and shape interpolation. Furthermore, we also show that including an estimation of the underlying density of the surface eases the learning process. By exploiting the potential of this architecture, we can train our model requiring only a sparse set of ground truth correspondences ($10\sim20\%$ of the total points). The proposed model and the analysis that we perform pave the way for future exploration of transformer-based architectures for registration and matching applications. Qualitative and quantitative evaluations demonstrate that our pipeline outperforms state-of-the-art methods for deformable and unordered 3D data registration on different datasets and scenarios.
[ { "created": "Fri, 25 Jun 2021 15:02:30 GMT", "version": "v1" }, { "created": "Mon, 28 Jun 2021 07:56:20 GMT", "version": "v2" } ]
2021-06-29
[ [ "Trappolini", "Giovanni", "" ], [ "Cosmo", "Luca", "" ], [ "Moschella", "Luca", "" ], [ "Marin", "Riccardo", "" ], [ "Melzi", "Simone", "" ], [ "Rodolà", "Emanuele", "" ] ]
In this paper, we propose a transformer-based procedure for the efficient registration of non-rigid 3D point clouds. The proposed approach is data-driven and adopts for the first time the transformer architecture in the registration task. Our method is general and applies to different settings. Given a fixed template with some desired properties (e.g. skinning weights or other animation cues), we can register raw acquired data to it, thereby transferring all the template properties to the input geometry. Alternatively, given a pair of shapes, our method can register the first onto the second (or vice-versa), obtaining a high-quality dense correspondence between the two. In both contexts, the quality of our results enables us to target real applications such as texture transfer and shape interpolation. Furthermore, we also show that including an estimation of the underlying density of the surface eases the learning process. By exploiting the potential of this architecture, we can train our model requiring only a sparse set of ground truth correspondences ($10\sim20\%$ of the total points). The proposed model and the analysis that we perform pave the way for future exploration of transformer-based architectures for registration and matching applications. Qualitative and quantitative evaluations demonstrate that our pipeline outperforms state-of-the-art methods for deformable and unordered 3D data registration on different datasets and scenarios.
1801.07743
Pedro Saleiro
Pedro Saleiro
Entity Retrieval and Text Mining for Online Reputation Monitoring
PhD Thesis
null
null
null
cs.IR cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Online Reputation Monitoring (ORM) is concerned with the use of computational tools to measure the reputation of entities online, such as politicians or companies. In practice, current ORM methods are constrained to the generation of data analytics reports, which aggregate statistics of popularity and sentiment on social media. We argue that this format is too restrictive as end users often like to have the flexibility to search for entity-centric information that is not available in predefined charts. As such, we propose the inclusion of entity retrieval capabilities as a first step towards the extension of current ORM capabilities. However, an entity's reputation is also influenced by the entity's relationships with other entities. Therefore, we address the problem of Entity-Relationship (E-R) retrieval in which the goal is to search for multiple connected entities. This is a challenging problem which traditional entity search systems cannot cope with. Besides E-R retrieval we also believe ORM would benefit of text-based entity-centric prediction capabilities, such as predicting entity popularity on social media based on news events or the outcome of political surveys. However, none of these tasks can provide useful results if there is no effective entity disambiguation and sentiment analysis tailored to the context of ORM. Consequently, this thesis address two computational problems in Online Reputation Monitoring: Entity Retrieval and Text Mining. We researched and developed methods to extract, retrieve and predict entity-centric information spread across the Web.
[ { "created": "Tue, 23 Jan 2018 19:36:29 GMT", "version": "v1" } ]
2018-01-25
[ [ "Saleiro", "Pedro", "" ] ]
Online Reputation Monitoring (ORM) is concerned with the use of computational tools to measure the reputation of entities online, such as politicians or companies. In practice, current ORM methods are constrained to the generation of data analytics reports, which aggregate statistics of popularity and sentiment on social media. We argue that this format is too restrictive as end users often like to have the flexibility to search for entity-centric information that is not available in predefined charts. As such, we propose the inclusion of entity retrieval capabilities as a first step towards the extension of current ORM capabilities. However, an entity's reputation is also influenced by the entity's relationships with other entities. Therefore, we address the problem of Entity-Relationship (E-R) retrieval in which the goal is to search for multiple connected entities. This is a challenging problem which traditional entity search systems cannot cope with. Besides E-R retrieval we also believe ORM would benefit of text-based entity-centric prediction capabilities, such as predicting entity popularity on social media based on news events or the outcome of political surveys. However, none of these tasks can provide useful results if there is no effective entity disambiguation and sentiment analysis tailored to the context of ORM. Consequently, this thesis address two computational problems in Online Reputation Monitoring: Entity Retrieval and Text Mining. We researched and developed methods to extract, retrieve and predict entity-centric information spread across the Web.
1811.08203
Noveen Sachdeva
Noveen Sachdeva, Kartik Gupta, Vikram Pudi
Attentive Neural Architecture Incorporating Song Features For Music Recommendation
Accepted as a paper at the 12th ACM Conference on Recommender Systems (RecSys 18)
12th ACM Conference on Recommender Systems (RecSys '18). ACM (2018) 417-421
10.1145/3240323.3240397
null
cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Recommender Systems are an integral part of music sharing platforms. Often the aim of these systems is to increase the time, the user spends on the platform and hence having a high commercial value. The systems which aim at increasing the average time a user spends on the platform often need to recommend songs which the user might want to listen to next at each point in time. This is different from recommendation systems which try to predict the item which might be of interest to the user at some point in the user lifetime but not necessarily in the very near future. Prediction of the next song the user might like requires some kind of modeling of the user interests at the given point of time. Attentive neural networks have been exploiting the sequence in which the items were selected by the user to model the implicit short-term interests of the user for the task of next item prediction, however we feel that the features of the songs occurring in the sequence could also convey some important information about the short-term user interest which only the items cannot. In this direction, we propose a novel attentive neural architecture which in addition to the sequence of items selected by the user, uses the features of these items to better learn the user short-term preferences and recommend the next song to the user.
[ { "created": "Tue, 20 Nov 2018 12:10:06 GMT", "version": "v1" } ]
2018-11-21
[ [ "Sachdeva", "Noveen", "" ], [ "Gupta", "Kartik", "" ], [ "Pudi", "Vikram", "" ] ]
Recommender Systems are an integral part of music sharing platforms. Often the aim of these systems is to increase the time, the user spends on the platform and hence having a high commercial value. The systems which aim at increasing the average time a user spends on the platform often need to recommend songs which the user might want to listen to next at each point in time. This is different from recommendation systems which try to predict the item which might be of interest to the user at some point in the user lifetime but not necessarily in the very near future. Prediction of the next song the user might like requires some kind of modeling of the user interests at the given point of time. Attentive neural networks have been exploiting the sequence in which the items were selected by the user to model the implicit short-term interests of the user for the task of next item prediction, however we feel that the features of the songs occurring in the sequence could also convey some important information about the short-term user interest which only the items cannot. In this direction, we propose a novel attentive neural architecture which in addition to the sequence of items selected by the user, uses the features of these items to better learn the user short-term preferences and recommend the next song to the user.
0710.2139
Ashkan Aazami
Ashkan Aazami, Michael D. Stilp
Approximation algorithms and hardness for domination with propagation
null
null
null
null
cs.CC cs.DM
null
The power dominating set (PDS) problem is the following extension of the well-known dominating set problem: find a smallest-size set of nodes $S$ that power dominates all the nodes, where a node $v$ is power dominated if (1) $v$ is in $S$ or $v$ has a neighbor in $S$, or (2) $v$ has a neighbor $w$ such that $w$ and all of its neighbors except $v$ are power dominated. We show a hardness of approximation threshold of $2^{\log^{1-\epsilon}{n}}$ in contrast to the logarithmic hardness for the dominating set problem. We give an $O(\sqrt{n})$ approximation algorithm for planar graphs, and show that our methods cannot improve on this approximation guarantee. Finally, we initiate the study of PDS on directed graphs, and show the same hardness threshold of $2^{\log^{1-\epsilon}{n}}$ for directed \emph{acyclic} graphs. Also we show that the directed PDS problem can be solved optimally in linear time if the underlying undirected graph has bounded tree-width.
[ { "created": "Wed, 10 Oct 2007 23:30:49 GMT", "version": "v1" } ]
2007-10-12
[ [ "Aazami", "Ashkan", "" ], [ "Stilp", "Michael D.", "" ] ]
The power dominating set (PDS) problem is the following extension of the well-known dominating set problem: find a smallest-size set of nodes $S$ that power dominates all the nodes, where a node $v$ is power dominated if (1) $v$ is in $S$ or $v$ has a neighbor in $S$, or (2) $v$ has a neighbor $w$ such that $w$ and all of its neighbors except $v$ are power dominated. We show a hardness of approximation threshold of $2^{\log^{1-\epsilon}{n}}$ in contrast to the logarithmic hardness for the dominating set problem. We give an $O(\sqrt{n})$ approximation algorithm for planar graphs, and show that our methods cannot improve on this approximation guarantee. Finally, we initiate the study of PDS on directed graphs, and show the same hardness threshold of $2^{\log^{1-\epsilon}{n}}$ for directed \emph{acyclic} graphs. Also we show that the directed PDS problem can be solved optimally in linear time if the underlying undirected graph has bounded tree-width.
2110.06750
Darius Sas
Darius Sas, Ilaria Pigazzini, Paris Avgeriou, Francesca Arcelli Fontana
The perception of Architectural Smells in industrial practice
Submitted and accepted to IEEE Software special issue on Technical Debt. This is a preprint
null
10.1109/MS.2021.3103664
null
cs.SE
http://creativecommons.org/licenses/by-nc-nd/4.0/
Architectural Technical Debt (ATD) is considered as the most significant type of TD in industrial practice. In this study, we interview 21 software engineers and architects to investigate a specific type of ATD, namely architectural smells (AS). Our goal is to understand the phenomenon of AS better and support practitioners to better manage it and researchers to offer relevant support. The findings of this study provide insights on how practitioners perceive AS and how they introduce them, the maintenance and evolution issues they experienced and associated to the presence of AS, and what practices and tools they adopt to manage AS.
[ { "created": "Wed, 13 Oct 2021 14:29:31 GMT", "version": "v1" } ]
2022-03-17
[ [ "Sas", "Darius", "" ], [ "Pigazzini", "Ilaria", "" ], [ "Avgeriou", "Paris", "" ], [ "Fontana", "Francesca Arcelli", "" ] ]
Architectural Technical Debt (ATD) is considered as the most significant type of TD in industrial practice. In this study, we interview 21 software engineers and architects to investigate a specific type of ATD, namely architectural smells (AS). Our goal is to understand the phenomenon of AS better and support practitioners to better manage it and researchers to offer relevant support. The findings of this study provide insights on how practitioners perceive AS and how they introduce them, the maintenance and evolution issues they experienced and associated to the presence of AS, and what practices and tools they adopt to manage AS.
2008.09645
Sheng Liu
Sheng Liu, Zuo-Jun Max Shen, Xiang Ji
Urban Bike Lane Planning with Bike Trajectories: Models, Algorithms, and a Real-World Case Study
null
null
null
null
cs.AI cs.CE math.OC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We study an urban bike lane planning problem based on the fine-grained bike trajectory data, which is made available by smart city infrastructure such as bike-sharing systems. The key decision is where to build bike lanes in the existing road network. As bike-sharing systems become widespread in the metropolitan areas over the world, bike lanes are being planned and constructed by many municipal governments to promote cycling and protect cyclists. Traditional bike lane planning approaches often rely on surveys and heuristics. We develop a general and novel optimization framework to guide the bike lane planning from bike trajectories. We formalize the bike lane planning problem in view of the cyclists' utility functions and derive an integer optimization model to maximize the utility. To capture cyclists' route choices, we develop a bilevel program based on the Multinomial Logit model. We derive structural properties about the base model and prove that the Lagrangian dual of the bike lane planning model is polynomial-time solvable. Furthermore, we reformulate the route choice based planning model as a mixed integer linear program using a linear approximation scheme. We develop tractable formulations and efficient algorithms to solve the large-scale optimization problem. Via a real-world case study with a city government, we demonstrate the efficiency of the proposed algorithms and quantify the trade-off between the coverage of bike trips and continuity of bike lanes. We show how the network topology evolves according to the utility functions and highlight the importance of understanding cyclists' route choices. The proposed framework drives the data-driven urban planning scheme in smart city operations management.
[ { "created": "Fri, 21 Aug 2020 18:46:51 GMT", "version": "v1" } ]
2020-08-25
[ [ "Liu", "Sheng", "" ], [ "Shen", "Zuo-Jun Max", "" ], [ "Ji", "Xiang", "" ] ]
We study an urban bike lane planning problem based on the fine-grained bike trajectory data, which is made available by smart city infrastructure such as bike-sharing systems. The key decision is where to build bike lanes in the existing road network. As bike-sharing systems become widespread in the metropolitan areas over the world, bike lanes are being planned and constructed by many municipal governments to promote cycling and protect cyclists. Traditional bike lane planning approaches often rely on surveys and heuristics. We develop a general and novel optimization framework to guide the bike lane planning from bike trajectories. We formalize the bike lane planning problem in view of the cyclists' utility functions and derive an integer optimization model to maximize the utility. To capture cyclists' route choices, we develop a bilevel program based on the Multinomial Logit model. We derive structural properties about the base model and prove that the Lagrangian dual of the bike lane planning model is polynomial-time solvable. Furthermore, we reformulate the route choice based planning model as a mixed integer linear program using a linear approximation scheme. We develop tractable formulations and efficient algorithms to solve the large-scale optimization problem. Via a real-world case study with a city government, we demonstrate the efficiency of the proposed algorithms and quantify the trade-off between the coverage of bike trips and continuity of bike lanes. We show how the network topology evolves according to the utility functions and highlight the importance of understanding cyclists' route choices. The proposed framework drives the data-driven urban planning scheme in smart city operations management.
1212.3879
EPTCS
Jurriaan Rot (LIACS - Leiden University, The Netherlands), Irina M\u{a}riuca As\u{a}voae (Alexandru Ioan Cuza University, Romania), Frank de Boer (Centrum Wiskunde en Informatica (CWI), The Netherlands), Marcello M. Bonsangue (LIACS - Leiden University, The Netherlands), Dorel Lucanu (Alexandru Ioan Cuza University, Romania)
Interacting via the Heap in the Presence of Recursion
In Proceedings ICE 2012, arXiv:1212.3458
EPTCS 104, 2012, pp. 99-113
10.4204/EPTCS.104.9
null
cs.PL cs.LO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Almost all modern imperative programming languages include operations for dynamically manipulating the heap, for example by allocating and deallocating objects, and by updating reference fields. In the presence of recursive procedures and local variables the interactions of a program with the heap can become rather complex, as an unbounded number of objects can be allocated either on the call stack using local variables, or, anonymously, on the heap using reference fields. As such a static analysis is, in general, undecidable. In this paper we study the verification of recursive programs with unbounded allocation of objects, in a simple imperative language for heap manipulation. We present an improved semantics for this language, using an abstraction that is precise. For any program with a bounded visible heap, meaning that the number of objects reachable from variables at any point of execution is bounded, this abstraction is a finitary representation of its behaviour, even though an unbounded number of objects can appear in the state. As a consequence, for such programs model checking is decidable. Finally we introduce a specification language for temporal properties of the heap, and discuss model checking these properties against heap-manipulating programs.
[ { "created": "Mon, 17 Dec 2012 03:42:47 GMT", "version": "v1" } ]
2012-12-18
[ [ "Rot", "Jurriaan", "", "LIACS - Leiden University, The Netherlands" ], [ "Asăvoae", "Irina Măriuca", "", "Alexandru Ioan Cuza University, Romania" ], [ "de Boer", "Frank", "", "Centrum Wiskunde en Informatica" ], [ "Bonsangue", "Marcello M.", "", "LIACS - Leiden University, The Netherlands" ], [ "Lucanu", "Dorel", "", "Alexandru Ioan Cuza University, Romania" ] ]
Almost all modern imperative programming languages include operations for dynamically manipulating the heap, for example by allocating and deallocating objects, and by updating reference fields. In the presence of recursive procedures and local variables the interactions of a program with the heap can become rather complex, as an unbounded number of objects can be allocated either on the call stack using local variables, or, anonymously, on the heap using reference fields. As such a static analysis is, in general, undecidable. In this paper we study the verification of recursive programs with unbounded allocation of objects, in a simple imperative language for heap manipulation. We present an improved semantics for this language, using an abstraction that is precise. For any program with a bounded visible heap, meaning that the number of objects reachable from variables at any point of execution is bounded, this abstraction is a finitary representation of its behaviour, even though an unbounded number of objects can appear in the state. As a consequence, for such programs model checking is decidable. Finally we introduce a specification language for temporal properties of the heap, and discuss model checking these properties against heap-manipulating programs.
1903.00565
Behnam Askarian
Fatemeh Sadat Tabei and Behnam Askarian
Effect of Proxy Nodes on the Performance of TCP-Based Transport Layer Protocols in Wireless Sensor Networks
null
null
null
null
cs.NI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Wireless Sensor Networks have recently attracted many researchers attentions due to their wide range of applications. Even though a plethora of studies have been carried out on characteristics, special conditions, and various aspects of WSNs, transport protocol which is compatible with conditions of Wireless Sensor Networks has not been considerably addressed. Wireless Sensor Networks have limitations such as storage space, energy resources, and wireless communication issues. Accordingly, widely-used transport protocols like Transmission Control Protocol (TCP) may not enjoy sufficient efficiency in such networks. In this paper, we study the characteristics of WSNs leading to design transport layer protocol for WSNs and aim at evaluating the efficiency of TCP and its dependent protocols (TCP variables), which are introduced to wireless networks. We propose to employ proxy nodes near sinks to improve the performance of transport layer. Our NS-2 simulation results indicate that throughput and packet delivery ratio are improved from 20 to 50 percent after employing proxy nodes, while the average message delay is almost increased twice.
[ { "created": "Fri, 1 Mar 2019 22:31:03 GMT", "version": "v1" } ]
2019-03-05
[ [ "Tabei", "Fatemeh Sadat", "" ], [ "Askarian", "Behnam", "" ] ]
Wireless Sensor Networks have recently attracted many researchers attentions due to their wide range of applications. Even though a plethora of studies have been carried out on characteristics, special conditions, and various aspects of WSNs, transport protocol which is compatible with conditions of Wireless Sensor Networks has not been considerably addressed. Wireless Sensor Networks have limitations such as storage space, energy resources, and wireless communication issues. Accordingly, widely-used transport protocols like Transmission Control Protocol (TCP) may not enjoy sufficient efficiency in such networks. In this paper, we study the characteristics of WSNs leading to design transport layer protocol for WSNs and aim at evaluating the efficiency of TCP and its dependent protocols (TCP variables), which are introduced to wireless networks. We propose to employ proxy nodes near sinks to improve the performance of transport layer. Our NS-2 simulation results indicate that throughput and packet delivery ratio are improved from 20 to 50 percent after employing proxy nodes, while the average message delay is almost increased twice.
2112.13808
Jamin Shin
Jamin Shin, Juneyoung Park
Pedagogical Word Recommendation: A novel task and dataset on personalized vocabulary acquisition for L2 learners
null
null
null
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
When learning a second language (L2), one of the most important but tedious components that often demoralizes students with its ineffectiveness and inefficiency is vocabulary acquisition, or more simply put, memorizing words. In light of such, a personalized and educational vocabulary recommendation system that traces a learner's vocabulary knowledge state would have an immense learning impact as it could resolve both issues. Therefore, in this paper, we propose and release data for a novel task called Pedagogical Word Recommendation (PWR). The main goal of PWR is to predict whether a given learner knows a given word based on other words the learner has already seen. To elaborate, we collect this data via an Intelligent Tutoring System (ITS) that is serviced to ~1M L2 learners who study for the standardized English exam, TOEIC. As a feature of this ITS, students can directly indicate words they do not know from the questions they solved to create wordbooks. Finally, we report the evaluation results of a Neural Collaborative Filtering approach along with an exploratory data analysis and discuss the impact and efficacy of this dataset as a baseline for future studies on this task.
[ { "created": "Mon, 27 Dec 2021 17:52:48 GMT", "version": "v1" }, { "created": "Tue, 28 Dec 2021 04:52:26 GMT", "version": "v2" } ]
2021-12-30
[ [ "Shin", "Jamin", "" ], [ "Park", "Juneyoung", "" ] ]
When learning a second language (L2), one of the most important but tedious components that often demoralizes students with its ineffectiveness and inefficiency is vocabulary acquisition, or more simply put, memorizing words. In light of such, a personalized and educational vocabulary recommendation system that traces a learner's vocabulary knowledge state would have an immense learning impact as it could resolve both issues. Therefore, in this paper, we propose and release data for a novel task called Pedagogical Word Recommendation (PWR). The main goal of PWR is to predict whether a given learner knows a given word based on other words the learner has already seen. To elaborate, we collect this data via an Intelligent Tutoring System (ITS) that is serviced to ~1M L2 learners who study for the standardized English exam, TOEIC. As a feature of this ITS, students can directly indicate words they do not know from the questions they solved to create wordbooks. Finally, we report the evaluation results of a Neural Collaborative Filtering approach along with an exploratory data analysis and discuss the impact and efficacy of this dataset as a baseline for future studies on this task.
2403.01649
Steven Bellovin
Susan Landau, James X. Dempsey, Ece Kamar, Steven M. Bellovin
Recommendations for Government Development and Use of Advanced Automated Systems to Make Decisions about Individuals
null
null
null
null
cs.CY cs.AI
http://creativecommons.org/licenses/by-nc-sa/4.0/
Contestability -- the ability to effectively challenge a decision -- is critical to the implementation of fairness. In the context of governmental decision making about individuals, contestability is often constitutionally required as an element of due process; specific procedures may be required by state or federal law relevant to a particular program. In addition, contestability can be a valuable way to discover systemic errors, contributing to ongoing assessments and system improvement. On January 24-25, 2024, with support from the National Science Foundation and the William and Flora Hewlett Foundation, we convened a diverse group of government officials, representatives of leading technology companies, technology and policy experts from academia and the non-profit sector, advocates, and stakeholders for a workshop on advanced automated decision making, contestability, and the law. Informed by the workshop's rich and wide-ranging discussion, we offer these recommendations. A full report summarizing the discussion is in preparation.
[ { "created": "Mon, 4 Mar 2024 00:03:00 GMT", "version": "v1" } ]
2024-03-05
[ [ "Landau", "Susan", "" ], [ "Dempsey", "James X.", "" ], [ "Kamar", "Ece", "" ], [ "Bellovin", "Steven M.", "" ] ]
Contestability -- the ability to effectively challenge a decision -- is critical to the implementation of fairness. In the context of governmental decision making about individuals, contestability is often constitutionally required as an element of due process; specific procedures may be required by state or federal law relevant to a particular program. In addition, contestability can be a valuable way to discover systemic errors, contributing to ongoing assessments and system improvement. On January 24-25, 2024, with support from the National Science Foundation and the William and Flora Hewlett Foundation, we convened a diverse group of government officials, representatives of leading technology companies, technology and policy experts from academia and the non-profit sector, advocates, and stakeholders for a workshop on advanced automated decision making, contestability, and the law. Informed by the workshop's rich and wide-ranging discussion, we offer these recommendations. A full report summarizing the discussion is in preparation.
2108.09956
Deborah Amos Phiri
Deborah Amos Phiri and Chipo Kanjo
Policy-Practice Contradiction: Case of Cloud Computing Adoption in the Malawi Health Sector
In proceedings of the 1st Virtual Conference on Implications of Information and Digital Technologies for Development, 2021
null
null
null
cs.CY
http://creativecommons.org/licenses/by-nc-sa/4.0/
This paper examines the dynamics of policy implementation and how policy contradicts reality on the ground when it comes to practice. The paper finds that despite having well-laid out policy; the actual practice is contrary. Taking data storage policy within the Ministry of Health in Malawi as a case study, the paper highlights that the contextual realities of where Ministry of Health (MoH) data is stored depends on a number of Technology-Organizational-Environmental (TOE) factors. In the wake of cloud computing; some of these factors act as causative factors for data to be stored in the cloud; contradicting the data storage policy.
[ { "created": "Mon, 23 Aug 2021 05:59:46 GMT", "version": "v1" } ]
2021-08-24
[ [ "Phiri", "Deborah Amos", "" ], [ "Kanjo", "Chipo", "" ] ]
This paper examines the dynamics of policy implementation and how policy contradicts reality on the ground when it comes to practice. The paper finds that despite having well-laid out policy; the actual practice is contrary. Taking data storage policy within the Ministry of Health in Malawi as a case study, the paper highlights that the contextual realities of where Ministry of Health (MoH) data is stored depends on a number of Technology-Organizational-Environmental (TOE) factors. In the wake of cloud computing; some of these factors act as causative factors for data to be stored in the cloud; contradicting the data storage policy.
1403.5468
Jian-Jun Shu
Jian-Jun Shu and Qi-Wen Wang
Beyond Parrondo's paradox
null
Scientific Reports, Vol. 4, No. 4244, pp. 1-9, 2014
10.1038/srep04244
null
cs.GT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The Parrondo's paradox is a counterintuitive phenomenon where individually-losing strategies can be combined in producing a winning expectation. In this paper, the issues surrounding the Parrondo's paradox are investigated. The focus is lying on testifying whether the same paradoxical effect can be reproduced by using a simple capital dependent game. The paradoxical effect generated by the Parrondo's paradox can be explained by placing all the parameters in one probability space. Based on this framework, it is able to generate other possible paradoxical effects by manipulating the parameters in the probability space.
[ { "created": "Fri, 21 Mar 2014 14:10:46 GMT", "version": "v1" } ]
2014-03-24
[ [ "Shu", "Jian-Jun", "" ], [ "Wang", "Qi-Wen", "" ] ]
The Parrondo's paradox is a counterintuitive phenomenon where individually-losing strategies can be combined in producing a winning expectation. In this paper, the issues surrounding the Parrondo's paradox are investigated. The focus is lying on testifying whether the same paradoxical effect can be reproduced by using a simple capital dependent game. The paradoxical effect generated by the Parrondo's paradox can be explained by placing all the parameters in one probability space. Based on this framework, it is able to generate other possible paradoxical effects by manipulating the parameters in the probability space.
1810.03120
Andrzej Pelc
Andrzej Pelc, Ram Narayan Yadav
Using Time to Break Symmetry: Universal Deterministic Anonymous Rendezvous
null
null
null
null
cs.DC
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Two anonymous mobile agents navigate synchronously in an anonymous graph and have to meet at a node, using a deterministic algorithm. This is a symmetry breaking task called rendezvous, equivalent to the fundamental task of leader election between the agents. When is this feasible in a completely anonymous environment? It is known that agents can always meet if their initial positions are nonsymmetric, and that if they are symmetric and agents start simultaneously then rendezvous is impossible. What happens for symmetric initial positions with non-simultaneous start? Can symmetry between the agents be broken by the delay between their starting times? In order to answer these questions, we consider {\em space-time initial configurations} (abbreviated by STIC). A STIC is formalized as $[(u,v),\delta]$, where $u$ and $v$ are initial nodes of the agents in some graph and $\delta$ is a non-negative integer that represents the difference between their starting times. A STIC is {\em feasible} if there exists a deterministic algorithm, even dedicated to this particular STIC, which accomplishes rendezvous for it. Our main result is a characterization of all feasible STICs and the design of a universal deterministic algorithm that accomplishes rendezvous for all of them without {\em any } a priori knowledge of the agents. Thus, as far as feasibility is concerned, we completely solve the problem of symmetry breaking between two anonymous agents in anonymous graphs. Moreover, we show that such a universal algorithm cannot work for all feasible STICs in time polynomial in the initial distance between the agents.
[ { "created": "Sun, 7 Oct 2018 11:16:59 GMT", "version": "v1" } ]
2018-10-09
[ [ "Pelc", "Andrzej", "" ], [ "Yadav", "Ram Narayan", "" ] ]
Two anonymous mobile agents navigate synchronously in an anonymous graph and have to meet at a node, using a deterministic algorithm. This is a symmetry breaking task called rendezvous, equivalent to the fundamental task of leader election between the agents. When is this feasible in a completely anonymous environment? It is known that agents can always meet if their initial positions are nonsymmetric, and that if they are symmetric and agents start simultaneously then rendezvous is impossible. What happens for symmetric initial positions with non-simultaneous start? Can symmetry between the agents be broken by the delay between their starting times? In order to answer these questions, we consider {\em space-time initial configurations} (abbreviated by STIC). A STIC is formalized as $[(u,v),\delta]$, where $u$ and $v$ are initial nodes of the agents in some graph and $\delta$ is a non-negative integer that represents the difference between their starting times. A STIC is {\em feasible} if there exists a deterministic algorithm, even dedicated to this particular STIC, which accomplishes rendezvous for it. Our main result is a characterization of all feasible STICs and the design of a universal deterministic algorithm that accomplishes rendezvous for all of them without {\em any } a priori knowledge of the agents. Thus, as far as feasibility is concerned, we completely solve the problem of symmetry breaking between two anonymous agents in anonymous graphs. Moreover, we show that such a universal algorithm cannot work for all feasible STICs in time polynomial in the initial distance between the agents.
2405.03724
Junxiang Wang
Junxiang Wang and Liang Zhao
GraphSL: An Open-Source Library for Graph Source Localization Approaches and Benchmark Datasets
null
null
null
null
cs.LG cs.SI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We introduce GraphSL, a new library for studying the graph source localization problem. graph diffusion and graph source localization are inverse problems in nature: graph diffusion predicts information diffusions from information sources, while graph source localization predicts information sources from information diffusions. GraphSL facilitates the exploration of various graph diffusion models for simulating information diffusions and enables the evaluation of cutting-edge source localization approaches on established benchmark datasets. The source code of GraphSL is made available at Github Repository (https://github.com/xianggebenben/GraphSL). Bug reports and feedback can be directed to the Github issues page (https://github.com/xianggebenben/GraphSL/issues).
[ { "created": "Mon, 6 May 2024 04:00:00 GMT", "version": "v1" }, { "created": "Sun, 28 Jul 2024 17:34:22 GMT", "version": "v2" } ]
2024-07-30
[ [ "Wang", "Junxiang", "" ], [ "Zhao", "Liang", "" ] ]
We introduce GraphSL, a new library for studying the graph source localization problem. graph diffusion and graph source localization are inverse problems in nature: graph diffusion predicts information diffusions from information sources, while graph source localization predicts information sources from information diffusions. GraphSL facilitates the exploration of various graph diffusion models for simulating information diffusions and enables the evaluation of cutting-edge source localization approaches on established benchmark datasets. The source code of GraphSL is made available at Github Repository (https://github.com/xianggebenben/GraphSL). Bug reports and feedback can be directed to the Github issues page (https://github.com/xianggebenben/GraphSL/issues).
1612.03849
Gabriele Oliva
Gabriele Oliva, Andrea Gasparri, Adriano Fagiolini, and Christoforos N. Hadjicostis
Distributed and Proximity-Constrained C-Means for Discrete Coverage Control
To appear in the 56th IEEE Conference on Decision and Control, to be held in Melbourne, Australia, December 12-15, 2017
null
null
null
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In this paper we present a novel distributed coverage control framework for a network of mobile agents, in charge of covering a finite set of points of interest (PoI), such as people in danger, geographically dispersed equipment or environmental landmarks. The proposed algorithm is inspired by C-Means, an unsupervised learning algorithm originally proposed for non-exclusive clustering and for identification of cluster centroids from a set of observations. To cope with the agents' limited sensing range and avoid infeasible coverage solutions, traditional C-Means needs to be enhanced with proximity constraints, ensuring that each agent takes into account only neighboring PoIs. The proposed coverage control framework provides useful information concerning the ranking or importance of the different PoIs to the agents, which can be exploited in further application-dependent data fusion processes, patrolling, or disaster relief applications.
[ { "created": "Mon, 12 Dec 2016 18:54:07 GMT", "version": "v1" }, { "created": "Thu, 16 Mar 2017 14:35:51 GMT", "version": "v2" }, { "created": "Fri, 8 Sep 2017 09:30:10 GMT", "version": "v3" }, { "created": "Sat, 16 Sep 2017 18:38:16 GMT", "version": "v4" } ]
2017-09-19
[ [ "Oliva", "Gabriele", "" ], [ "Gasparri", "Andrea", "" ], [ "Fagiolini", "Adriano", "" ], [ "Hadjicostis", "Christoforos N.", "" ] ]
In this paper we present a novel distributed coverage control framework for a network of mobile agents, in charge of covering a finite set of points of interest (PoI), such as people in danger, geographically dispersed equipment or environmental landmarks. The proposed algorithm is inspired by C-Means, an unsupervised learning algorithm originally proposed for non-exclusive clustering and for identification of cluster centroids from a set of observations. To cope with the agents' limited sensing range and avoid infeasible coverage solutions, traditional C-Means needs to be enhanced with proximity constraints, ensuring that each agent takes into account only neighboring PoIs. The proposed coverage control framework provides useful information concerning the ranking or importance of the different PoIs to the agents, which can be exploited in further application-dependent data fusion processes, patrolling, or disaster relief applications.
1502.04662
Tim Althoff
Tim Althoff, Xin Luna Dong, Kevin Murphy, Safa Alai, Van Dang, Wei Zhang
TimeMachine: Timeline Generation for Knowledge-Base Entities
To appear at ACM SIGKDD KDD'15. 12pp, 7 fig. With appendix. Demo and other info available at http://cs.stanford.edu/~althoff/timemachine/
null
10.1145/2783258.2783325
null
cs.DB cs.IR
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
We present a method called TIMEMACHINE to generate a timeline of events and relations for entities in a knowledge base. For example for an actor, such a timeline should show the most important professional and personal milestones and relationships such as works, awards, collaborations, and family relationships. We develop three orthogonal timeline quality criteria that an ideal timeline should satisfy: (1) it shows events that are relevant to the entity; (2) it shows events that are temporally diverse, so they distribute along the time axis, avoiding visual crowding and allowing for easy user interaction, such as zooming in and out; and (3) it shows events that are content diverse, so they contain many different types of events (e.g., for an actor, it should show movies and marriages and awards, not just movies). We present an algorithm to generate such timelines for a given time period and screen size, based on submodular optimization and web-co-occurrence statistics with provable performance guarantees. A series of user studies using Mechanical Turk shows that all three quality criteria are crucial to produce quality timelines and that our algorithm significantly outperforms various baseline and state-of-the-art methods.
[ { "created": "Mon, 16 Feb 2015 18:53:01 GMT", "version": "v1" }, { "created": "Sat, 21 Feb 2015 07:02:11 GMT", "version": "v2" }, { "created": "Mon, 8 Jun 2015 20:39:26 GMT", "version": "v3" } ]
2015-06-10
[ [ "Althoff", "Tim", "" ], [ "Dong", "Xin Luna", "" ], [ "Murphy", "Kevin", "" ], [ "Alai", "Safa", "" ], [ "Dang", "Van", "" ], [ "Zhang", "Wei", "" ] ]
We present a method called TIMEMACHINE to generate a timeline of events and relations for entities in a knowledge base. For example for an actor, such a timeline should show the most important professional and personal milestones and relationships such as works, awards, collaborations, and family relationships. We develop three orthogonal timeline quality criteria that an ideal timeline should satisfy: (1) it shows events that are relevant to the entity; (2) it shows events that are temporally diverse, so they distribute along the time axis, avoiding visual crowding and allowing for easy user interaction, such as zooming in and out; and (3) it shows events that are content diverse, so they contain many different types of events (e.g., for an actor, it should show movies and marriages and awards, not just movies). We present an algorithm to generate such timelines for a given time period and screen size, based on submodular optimization and web-co-occurrence statistics with provable performance guarantees. A series of user studies using Mechanical Turk shows that all three quality criteria are crucial to produce quality timelines and that our algorithm significantly outperforms various baseline and state-of-the-art methods.
2107.03140
Bahram Sadeghi Bigham
Bahram Sadeghi Bigham
Minimum Constraint Removal Problem for Line Segments is NP-hard
null
Bigham, Bahram Sadeghi. "Minimum constraint removal problem for line segments is NP-hard." Discrete Mathematics, Algorithms and Applications (2022): 2250055
null
null
cs.CG cs.RO
http://creativecommons.org/licenses/by-nc-nd/4.0/
In the minimum constraint removal ($MCR$), there is no feasible path to move from the starting point towards the goal and, the minimum constraints should be removed in order to find a collision-free path. It has been proved that $MCR$ problem is $NP-hard$ when constraints have arbitrary shapes or even they are in shape of convex polygons. However, it has a simple linear solution when constraints are lines and the problem is open for other cases yet. In this paper, using a reduction from Subset Sum problem, in three steps, we show that the problem is NP-hard for both weighted and unweighted line segments.
[ { "created": "Wed, 7 Jul 2021 10:57:22 GMT", "version": "v1" } ]
2023-02-21
[ [ "Bigham", "Bahram Sadeghi", "" ] ]
In the minimum constraint removal ($MCR$), there is no feasible path to move from the starting point towards the goal and, the minimum constraints should be removed in order to find a collision-free path. It has been proved that $MCR$ problem is $NP-hard$ when constraints have arbitrary shapes or even they are in shape of convex polygons. However, it has a simple linear solution when constraints are lines and the problem is open for other cases yet. In this paper, using a reduction from Subset Sum problem, in three steps, we show that the problem is NP-hard for both weighted and unweighted line segments.
1904.05272
Tang Liu
Tang Liu and Daniela Tuninetti
Decentralized Pliable Index Coding
5 pages. To be presented at ISIT 2019
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper introduces the ${\it decentralized}$ Pliable Index CODing (PICOD) problem: a variant of the Index Coding (IC) problem, where a central transmitter serves ${\it pliable}$ users with message side information; here, pliable refers to the fact that a user is satisfied by decoding ${\it any}$ $t$ messages that are not in its side information set. In the decentralized PICOD, a central transmitter with knowledge of all messages is not present, and instead users share among themselves massages that can only depend on their local side information set. This paper characterizes the capacity of two classes of decentralized complete--$S$ PICOD$(t)$ problems with $m$ messages (where the set $S\subset[m]$ contains the sizes of the side information sets, and the number of users is $n=\sum_{s\in S}\binom{m}{s}$, with no two users having the same side information set): (i) the consecutive case: $S=[s_\min:s_\max]$ for some $0 \leq s_\min\leq s_\max \leq m-t$, and (ii) the complement-consecutive case: $S=[0:m-t]\backslash[s_\min:s_\max]$, for some $0 < s_\min\leq s_\max < m-t$. Interestingly, the optimal code-length for the decentralized PICOD in those cases is the same as for the classical (centralized) PICOD counterpart, except when the problem is no longer pliable, that is, it reduces to an IC problem where every user needs to decode all messages not in its side information set. Although the optimal code-length may be the same in both centralized and decentralized settings, the actual optimal codes are not. For the decentralized PICOD, sparse Maximum Distance Separable (MDS) codes and vector linear index codes are used (as opposed to scalar linear codes).
[ { "created": "Wed, 10 Apr 2019 16:18:57 GMT", "version": "v1" } ]
2019-04-11
[ [ "Liu", "Tang", "" ], [ "Tuninetti", "Daniela", "" ] ]
This paper introduces the ${\it decentralized}$ Pliable Index CODing (PICOD) problem: a variant of the Index Coding (IC) problem, where a central transmitter serves ${\it pliable}$ users with message side information; here, pliable refers to the fact that a user is satisfied by decoding ${\it any}$ $t$ messages that are not in its side information set. In the decentralized PICOD, a central transmitter with knowledge of all messages is not present, and instead users share among themselves massages that can only depend on their local side information set. This paper characterizes the capacity of two classes of decentralized complete--$S$ PICOD$(t)$ problems with $m$ messages (where the set $S\subset[m]$ contains the sizes of the side information sets, and the number of users is $n=\sum_{s\in S}\binom{m}{s}$, with no two users having the same side information set): (i) the consecutive case: $S=[s_\min:s_\max]$ for some $0 \leq s_\min\leq s_\max \leq m-t$, and (ii) the complement-consecutive case: $S=[0:m-t]\backslash[s_\min:s_\max]$, for some $0 < s_\min\leq s_\max < m-t$. Interestingly, the optimal code-length for the decentralized PICOD in those cases is the same as for the classical (centralized) PICOD counterpart, except when the problem is no longer pliable, that is, it reduces to an IC problem where every user needs to decode all messages not in its side information set. Although the optimal code-length may be the same in both centralized and decentralized settings, the actual optimal codes are not. For the decentralized PICOD, sparse Maximum Distance Separable (MDS) codes and vector linear index codes are used (as opposed to scalar linear codes).
2309.01398
Danqing Hu
Danqing Hu, Bing Liu, Xiaofeng Zhu, Xudong Lu, Nan Wu
Zero-shot information extraction from radiological reports using ChatGPT
null
null
10.1016/j.ijmedinf.2023.105321
null
cs.CL
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Electronic health records contain an enormous amount of valuable information, but many are recorded in free text. Information extraction is the strategy to transform the sequence of characters into structured data, which can be employed for secondary analysis. However, the traditional information extraction components, such as named entity recognition and relation extraction, require annotated data to optimize the model parameters, which has become one of the major bottlenecks in building information extraction systems. With the large language models achieving good performances on various downstream NLP tasks without parameter tuning, it becomes possible to use large language models for zero-shot information extraction. In this study, we aim to explore whether the most popular large language model, ChatGPT, can extract useful information from the radiological reports. We first design the prompt template for the interested information in the CT reports. Then, we generate the prompts by combining the prompt template with the CT reports as the inputs of ChatGPT to obtain the responses. A post-processing module is developed to transform the responses into structured extraction results. We conducted the experiments with 847 CT reports collected from Peking University Cancer Hospital. The experimental results indicate that ChatGPT can achieve competitive performances for some extraction tasks compared with the baseline information extraction system, but some limitations need to be further improved.
[ { "created": "Mon, 4 Sep 2023 07:00:26 GMT", "version": "v1" }, { "created": "Thu, 7 Sep 2023 01:36:08 GMT", "version": "v2" } ]
2024-01-03
[ [ "Hu", "Danqing", "" ], [ "Liu", "Bing", "" ], [ "Zhu", "Xiaofeng", "" ], [ "Lu", "Xudong", "" ], [ "Wu", "Nan", "" ] ]
Electronic health records contain an enormous amount of valuable information, but many are recorded in free text. Information extraction is the strategy to transform the sequence of characters into structured data, which can be employed for secondary analysis. However, the traditional information extraction components, such as named entity recognition and relation extraction, require annotated data to optimize the model parameters, which has become one of the major bottlenecks in building information extraction systems. With the large language models achieving good performances on various downstream NLP tasks without parameter tuning, it becomes possible to use large language models for zero-shot information extraction. In this study, we aim to explore whether the most popular large language model, ChatGPT, can extract useful information from the radiological reports. We first design the prompt template for the interested information in the CT reports. Then, we generate the prompts by combining the prompt template with the CT reports as the inputs of ChatGPT to obtain the responses. A post-processing module is developed to transform the responses into structured extraction results. We conducted the experiments with 847 CT reports collected from Peking University Cancer Hospital. The experimental results indicate that ChatGPT can achieve competitive performances for some extraction tasks compared with the baseline information extraction system, but some limitations need to be further improved.
2402.03413
Huiyu Duan
Xiongkuo Min, Huiyu Duan, Wei Sun, Yucheng Zhu, Guangtao Zhai
Perceptual Video Quality Assessment: A Survey
null
null
null
null
cs.MM cs.CV eess.IV
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Perceptual video quality assessment plays a vital role in the field of video processing due to the existence of quality degradations introduced in various stages of video signal acquisition, compression, transmission and display. With the advancement of internet communication and cloud service technology, video content and traffic are growing exponentially, which further emphasizes the requirement for accurate and rapid assessment of video quality. Therefore, numerous subjective and objective video quality assessment studies have been conducted over the past two decades for both generic videos and specific videos such as streaming, user-generated content (UGC), 3D, virtual and augmented reality (VR and AR), high frame rate (HFR), audio-visual, etc. This survey provides an up-to-date and comprehensive review of these video quality assessment studies. Specifically, we first review the subjective video quality assessment methodologies and databases, which are necessary for validating the performance of video quality metrics. Second, the objective video quality assessment algorithms for general purposes are surveyed and concluded according to the methodologies utilized in the quality measures. Third, we overview the objective video quality assessment measures for specific applications and emerging topics. Finally, the performances of the state-of-the-art video quality assessment measures are compared and analyzed. This survey provides a systematic overview of both classical works and recent progresses in the realm of video quality assessment, which can help other researchers quickly access the field and conduct relevant research.
[ { "created": "Mon, 5 Feb 2024 16:13:52 GMT", "version": "v1" } ]
2024-02-07
[ [ "Min", "Xiongkuo", "" ], [ "Duan", "Huiyu", "" ], [ "Sun", "Wei", "" ], [ "Zhu", "Yucheng", "" ], [ "Zhai", "Guangtao", "" ] ]
Perceptual video quality assessment plays a vital role in the field of video processing due to the existence of quality degradations introduced in various stages of video signal acquisition, compression, transmission and display. With the advancement of internet communication and cloud service technology, video content and traffic are growing exponentially, which further emphasizes the requirement for accurate and rapid assessment of video quality. Therefore, numerous subjective and objective video quality assessment studies have been conducted over the past two decades for both generic videos and specific videos such as streaming, user-generated content (UGC), 3D, virtual and augmented reality (VR and AR), high frame rate (HFR), audio-visual, etc. This survey provides an up-to-date and comprehensive review of these video quality assessment studies. Specifically, we first review the subjective video quality assessment methodologies and databases, which are necessary for validating the performance of video quality metrics. Second, the objective video quality assessment algorithms for general purposes are surveyed and concluded according to the methodologies utilized in the quality measures. Third, we overview the objective video quality assessment measures for specific applications and emerging topics. Finally, the performances of the state-of-the-art video quality assessment measures are compared and analyzed. This survey provides a systematic overview of both classical works and recent progresses in the realm of video quality assessment, which can help other researchers quickly access the field and conduct relevant research.
1605.06417
Yuan Jiang
Wei Shen, Yuan Jiang, Wenjing Gao, Dan Zeng, Xinggang Wang
Shape Recognition by Bag of Skeleton-associated Contour Parts
10 pages. Has been Accepted by Pattern Recognition Letters 2016
null
10.1007/978-3-662-45646-0_40
null
cs.CV
http://creativecommons.org/licenses/by-nc-sa/4.0/
Contour and skeleton are two complementary representations for shape recognition. However combining them in a principal way is nontrivial, as they are generally abstracted by different structures (closed string vs graph), respectively. This paper aims at addressing the shape recognition problem by combining contour and skeleton according to the correspondence between them. The correspondence provides a straightforward way to associate skeletal information with a shape contour. More specifically, we propose a new shape descriptor. named Skeleton-associated Shape Context (SSC), which captures the features of a contour fragment associated with skeletal information. Benefited from the association, the proposed shape descriptor provides the complementary geometric information from both contour and skeleton parts, including the spatial distribution and the thickness change along the shape part. To form a meaningful shape feature vector for an overall shape, the Bag of Features framework is applied to the SSC descriptors extracted from it. Finally, the shape feature vector is fed into a linear SVM classifier to recognize the shape. The encouraging experimental results demonstrate that the proposed way to combine contour and skeleton is effective for shape recognition, which achieves the state-of-the-art performances on several standard shape benchmarks.
[ { "created": "Fri, 20 May 2016 16:07:41 GMT", "version": "v1" } ]
2016-05-23
[ [ "Shen", "Wei", "" ], [ "Jiang", "Yuan", "" ], [ "Gao", "Wenjing", "" ], [ "Zeng", "Dan", "" ], [ "Wang", "Xinggang", "" ] ]
Contour and skeleton are two complementary representations for shape recognition. However combining them in a principal way is nontrivial, as they are generally abstracted by different structures (closed string vs graph), respectively. This paper aims at addressing the shape recognition problem by combining contour and skeleton according to the correspondence between them. The correspondence provides a straightforward way to associate skeletal information with a shape contour. More specifically, we propose a new shape descriptor. named Skeleton-associated Shape Context (SSC), which captures the features of a contour fragment associated with skeletal information. Benefited from the association, the proposed shape descriptor provides the complementary geometric information from both contour and skeleton parts, including the spatial distribution and the thickness change along the shape part. To form a meaningful shape feature vector for an overall shape, the Bag of Features framework is applied to the SSC descriptors extracted from it. Finally, the shape feature vector is fed into a linear SVM classifier to recognize the shape. The encouraging experimental results demonstrate that the proposed way to combine contour and skeleton is effective for shape recognition, which achieves the state-of-the-art performances on several standard shape benchmarks.
2306.00809
Emanuele Francazi
Emanuele Francazi, Aurelien Lucchi, Marco Baity-Jesi
Initial Guessing Bias: How Untrained Networks Favor Some Classes
We have added experiments on pre-trained models and various new results, including analysis in the limit of an infinite number of classes and an extension of the analysis to non-identically distributed classes. Additionally, we have slightly restructured the main paper to include more discussion on the practical implications of the phenomenon
null
null
null
cs.LG cond-mat.dis-nn stat.ML
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Understanding and controlling biasing effects in neural networks is crucial for ensuring accurate and fair model performance. In the context of classification problems, we provide a theoretical analysis demonstrating that the structure of a deep neural network (DNN) can condition the model to assign all predictions to the same class, even before the beginning of training, and in the absence of explicit biases. We prove that, besides dataset properties, the presence of this phenomenon, which we call \textit{Initial Guessing Bias} (IGB), is influenced by model choices including dataset preprocessing methods, and architectural decisions, such as activation functions, max-pooling layers, and network depth. Our analysis of IGB provides information for architecture selection and model initialization. We also highlight theoretical consequences, such as the breakdown of node-permutation symmetry, the violation of self-averaging and the non-trivial effects that depth has on the phenomenon.
[ { "created": "Thu, 1 Jun 2023 15:37:32 GMT", "version": "v1" }, { "created": "Wed, 1 Nov 2023 16:17:43 GMT", "version": "v2" }, { "created": "Mon, 12 Feb 2024 12:48:53 GMT", "version": "v3" }, { "created": "Thu, 13 Jun 2024 22:30:36 GMT", "version": "v4" } ]
2024-06-17
[ [ "Francazi", "Emanuele", "" ], [ "Lucchi", "Aurelien", "" ], [ "Baity-Jesi", "Marco", "" ] ]
Understanding and controlling biasing effects in neural networks is crucial for ensuring accurate and fair model performance. In the context of classification problems, we provide a theoretical analysis demonstrating that the structure of a deep neural network (DNN) can condition the model to assign all predictions to the same class, even before the beginning of training, and in the absence of explicit biases. We prove that, besides dataset properties, the presence of this phenomenon, which we call \textit{Initial Guessing Bias} (IGB), is influenced by model choices including dataset preprocessing methods, and architectural decisions, such as activation functions, max-pooling layers, and network depth. Our analysis of IGB provides information for architecture selection and model initialization. We also highlight theoretical consequences, such as the breakdown of node-permutation symmetry, the violation of self-averaging and the non-trivial effects that depth has on the phenomenon.
1509.05636
Seetha Ramaiah M
M. Seetha Ramaiah, Amitabha Mukerjee, Arindam Chakraborty, Sadbodh Sharma
Visual Generalized Coordinates
null
null
null
null
cs.RO cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
An open problem in robotics is that of using vision to identify a robot's own body and the world around it. Many models attempt to recover the traditional C-space parameters. Instead, we propose an alternative C-space by deriving generalized coordinates from $n$ images of the robot. We show that the space of such images is bijective to the motion space, so these images lie on a manifold $\mathcal{V}$ homeomorphic to the canonical C-space. We now approximate this manifold as a set of $n$ neighbourhood tangent spaces that result in a graph, which we call the Visual Roadmap (VRM). Given a new robot image, we perform inverse kinematics visually by interpolating between nearby images in the image space. Obstacles are projected onto the VRM in $O(n)$ time by superimposition of images, leading to the identification of collision poses. The edges joining the free nodes can now be checked with a visual local planner, and free-space motions computed in $O(nlogn)$ time. This enables us to plan paths in the image space for a robot manipulator with unknown link geometries, DOF, kinematics, obstacles, and camera pose. We sketch the proofs for the main theoretical ideas, identify the assumptions, and demonstrate the approach for both articulated and mobile robots. We also investigate the feasibility of the process by investigating various metrics and image sampling densities, and demonstrate it on simulated and real robots.
[ { "created": "Fri, 18 Sep 2015 14:17:57 GMT", "version": "v1" } ]
2015-09-21
[ [ "Ramaiah", "M. Seetha", "" ], [ "Mukerjee", "Amitabha", "" ], [ "Chakraborty", "Arindam", "" ], [ "Sharma", "Sadbodh", "" ] ]
An open problem in robotics is that of using vision to identify a robot's own body and the world around it. Many models attempt to recover the traditional C-space parameters. Instead, we propose an alternative C-space by deriving generalized coordinates from $n$ images of the robot. We show that the space of such images is bijective to the motion space, so these images lie on a manifold $\mathcal{V}$ homeomorphic to the canonical C-space. We now approximate this manifold as a set of $n$ neighbourhood tangent spaces that result in a graph, which we call the Visual Roadmap (VRM). Given a new robot image, we perform inverse kinematics visually by interpolating between nearby images in the image space. Obstacles are projected onto the VRM in $O(n)$ time by superimposition of images, leading to the identification of collision poses. The edges joining the free nodes can now be checked with a visual local planner, and free-space motions computed in $O(nlogn)$ time. This enables us to plan paths in the image space for a robot manipulator with unknown link geometries, DOF, kinematics, obstacles, and camera pose. We sketch the proofs for the main theoretical ideas, identify the assumptions, and demonstrate the approach for both articulated and mobile robots. We also investigate the feasibility of the process by investigating various metrics and image sampling densities, and demonstrate it on simulated and real robots.
1102.0454
Arnau Ramisa
Arnau Ramisa, David Aldavert, Shrihari Vasudevan, Ricardo Toledo, Ramon Lopez de Mantaras
Evaluation of Three Vision Based Object Perception Methods for a Mobile Robot
37 pages, 11 figures
null
null
IIIA research report 2011-01
cs.RO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper addresses object perception applied to mobile robotics. Being able to perceive semantically meaningful objects in unstructured environments is a key capability in order to make robots suitable to perform high-level tasks in home environments. However, finding a solution for this task is daunting: it requires the ability to handle the variability in image formation in a moving camera with tight time constraints. The paper brings to attention some of the issues with applying three state of the art object recognition and detection methods in a mobile robotics scenario, and proposes methods to deal with windowing/segmentation. Thus, this work aims at evaluating the state-of-the-art in object perception in an attempt to develop a lightweight solution for mobile robotics use/research in typical indoor settings.
[ { "created": "Wed, 2 Feb 2011 15:00:09 GMT", "version": "v1" } ]
2015-03-18
[ [ "Ramisa", "Arnau", "" ], [ "Aldavert", "David", "" ], [ "Vasudevan", "Shrihari", "" ], [ "Toledo", "Ricardo", "" ], [ "de Mantaras", "Ramon Lopez", "" ] ]
This paper addresses object perception applied to mobile robotics. Being able to perceive semantically meaningful objects in unstructured environments is a key capability in order to make robots suitable to perform high-level tasks in home environments. However, finding a solution for this task is daunting: it requires the ability to handle the variability in image formation in a moving camera with tight time constraints. The paper brings to attention some of the issues with applying three state of the art object recognition and detection methods in a mobile robotics scenario, and proposes methods to deal with windowing/segmentation. Thus, this work aims at evaluating the state-of-the-art in object perception in an attempt to develop a lightweight solution for mobile robotics use/research in typical indoor settings.
1607.06042
Vaneet Aggarwal
Mehdi Ashraphijuo and Vaneet Aggarwal and Xiaodong Wang
The DoF of Two-way Butterfly Networks
null
IEEE Communications Letters, Volume: 21, Issue: 10, Oct. 2017
10.1109/LCOMM.2017.2723364
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper studies the two-way butterfly network, a class of two-way four-unicast networks. We first show that bidirectional links do not increase the degrees of freedom for this network thus giving the first example for networks, to the best of our knowledge, where bidirectional links do not increase the degrees of freedom. Further, we see that sufficient caching at the relays or increasing the number of antennas in the relays can double the two-way degrees of freedom for butterfly network.
[ { "created": "Wed, 20 Jul 2016 17:44:49 GMT", "version": "v1" } ]
2017-11-03
[ [ "Ashraphijuo", "Mehdi", "" ], [ "Aggarwal", "Vaneet", "" ], [ "Wang", "Xiaodong", "" ] ]
This paper studies the two-way butterfly network, a class of two-way four-unicast networks. We first show that bidirectional links do not increase the degrees of freedom for this network thus giving the first example for networks, to the best of our knowledge, where bidirectional links do not increase the degrees of freedom. Further, we see that sufficient caching at the relays or increasing the number of antennas in the relays can double the two-way degrees of freedom for butterfly network.
1605.03518
Yang Huang
Yang Huang and Bruno Clerckx
Relaying Strategies for Wireless-Powered MIMO Relay Networks
Submitted for possible journal publication
null
null
null
cs.IT math.IT
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
This paper investigates relaying schemes in an amplify-and-forward multiple-input multiple-output relay network, where an energy-constrained relay harvests wireless power from the source information flow and can be further aided by an energy flow (EF) in the form of a wireless power transfer at the destination. However, the joint optimization of the relay matrix and the source precoder for the energy-flow-assisted (EFA) and the non-EFA (NEFA) schemes is intractable. The original rate maximization problem is transformed into an equivalent weighted mean square error minimization problem and optimized iteratively, where the global optimum of the nonconvex source precoder subproblem is achieved by semidefinite relaxation and rank reduction. The iterative algorithm finally converges. Then, the simplified EFA and NEFA schemes are proposed based on channel diagonalization, such that the matrices optimizations can be simplified to power optimizations. Closed-form solutions can be achieved. Simulation results reveal that the EFA schemes can outperform the NEFA schemes. Additionally, deploying more antennas at the relay increases the dimension of the signal space at the relay. Exploiting the additional dimension, the EF leakage in the information detecting block can be nearly separated from the information signal, such that the EF leakage can be amplified with a small coefficient.
[ { "created": "Wed, 11 May 2016 17:19:45 GMT", "version": "v1" } ]
2016-05-12
[ [ "Huang", "Yang", "" ], [ "Clerckx", "Bruno", "" ] ]
This paper investigates relaying schemes in an amplify-and-forward multiple-input multiple-output relay network, where an energy-constrained relay harvests wireless power from the source information flow and can be further aided by an energy flow (EF) in the form of a wireless power transfer at the destination. However, the joint optimization of the relay matrix and the source precoder for the energy-flow-assisted (EFA) and the non-EFA (NEFA) schemes is intractable. The original rate maximization problem is transformed into an equivalent weighted mean square error minimization problem and optimized iteratively, where the global optimum of the nonconvex source precoder subproblem is achieved by semidefinite relaxation and rank reduction. The iterative algorithm finally converges. Then, the simplified EFA and NEFA schemes are proposed based on channel diagonalization, such that the matrices optimizations can be simplified to power optimizations. Closed-form solutions can be achieved. Simulation results reveal that the EFA schemes can outperform the NEFA schemes. Additionally, deploying more antennas at the relay increases the dimension of the signal space at the relay. Exploiting the additional dimension, the EF leakage in the information detecting block can be nearly separated from the information signal, such that the EF leakage can be amplified with a small coefficient.
2402.15769
Zeming Dong
Zeming Dong, Qiang Hu, Xiaofei Xie, Maxime Cordy, Mike Papadakis, Jianjun Zhao
Importance Guided Data Augmentation for Neural-Based Code Understanding
null
null
null
null
cs.SE cs.AI
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Pre-trained code models lead the era of code intelligence. Many models have been designed with impressive performance recently. However, one important problem, data augmentation for code data that automatically helps developers prepare training data lacks study in the field of code learning. In this paper, we introduce a general data augmentation framework, GenCode, to enhance the training of code understanding models. GenCode follows a generation-and-selection paradigm to prepare useful training codes. Specifically, it uses code transformation techniques to generate new code candidates first and then selects important ones as the training data by importance metrics. To evaluate the effectiveness of GenCode with a general importance metric -- loss value, we conduct experiments on four code understanding tasks (e.g., code clone detection) and three pre-trained code models (e.g., CodeT5). Compared to the state-of-the-art (SOTA) code augmentation method, MixCode, GenCode produces code models with 2.92% higher accuracy and 4.90% robustness on average.
[ { "created": "Sat, 24 Feb 2024 08:57:12 GMT", "version": "v1" } ]
2024-02-27
[ [ "Dong", "Zeming", "" ], [ "Hu", "Qiang", "" ], [ "Xie", "Xiaofei", "" ], [ "Cordy", "Maxime", "" ], [ "Papadakis", "Mike", "" ], [ "Zhao", "Jianjun", "" ] ]
Pre-trained code models lead the era of code intelligence. Many models have been designed with impressive performance recently. However, one important problem, data augmentation for code data that automatically helps developers prepare training data lacks study in the field of code learning. In this paper, we introduce a general data augmentation framework, GenCode, to enhance the training of code understanding models. GenCode follows a generation-and-selection paradigm to prepare useful training codes. Specifically, it uses code transformation techniques to generate new code candidates first and then selects important ones as the training data by importance metrics. To evaluate the effectiveness of GenCode with a general importance metric -- loss value, we conduct experiments on four code understanding tasks (e.g., code clone detection) and three pre-trained code models (e.g., CodeT5). Compared to the state-of-the-art (SOTA) code augmentation method, MixCode, GenCode produces code models with 2.92% higher accuracy and 4.90% robustness on average.
1807.00602
Evgeniy Gryaznov
Evgeniy Gryaznov
Semantic Query Language for Temporal Genealogical Trees
null
null
null
null
cs.DB
http://creativecommons.org/licenses/by-sa/4.0/
Computers play a crucial role in modern ancestry management, they are used to collect, store, analyze, sort and display genealogical data. However, current applications do not take into account the kinship structure of a natural language. In this paper we propose a new domain-specific language KISP which is based on a formalization of English kinship system, for accessing and querying traditional genealogical trees. KISP is a dynamically typed LISP-like programming language with a rich set of features, such as kinship term reduction and temporal information expression. Our solution provides a user with a coherent genealogical framework that allows for a natural navigation over any traditional family tree.
[ { "created": "Mon, 2 Jul 2018 11:27:51 GMT", "version": "v1" } ]
2018-07-03
[ [ "Gryaznov", "Evgeniy", "" ] ]
Computers play a crucial role in modern ancestry management, they are used to collect, store, analyze, sort and display genealogical data. However, current applications do not take into account the kinship structure of a natural language. In this paper we propose a new domain-specific language KISP which is based on a formalization of English kinship system, for accessing and querying traditional genealogical trees. KISP is a dynamically typed LISP-like programming language with a rich set of features, such as kinship term reduction and temporal information expression. Our solution provides a user with a coherent genealogical framework that allows for a natural navigation over any traditional family tree.
2202.08870
Pat Morin
Prosenjit Bose, Pat Morin, and Saeed Odak
An Optimal Algorithm for Product Structure in Planar Graphs
14 pages, 5 figures
null
null
null
cs.DS math.CO
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
The \emph{Product Structure Theorem} for planar graphs (Dujmovi\'c et al.\ \emph{JACM}, \textbf{67}(4):22) states that any planar graph is contained in the strong product of a planar $3$-tree, a path, and a $3$-cycle. We give a simple linear-time algorithm for finding this decomposition as well as several related decompositions. This improves on the previous $O(n\log n)$ time algorithm (Morin.\ \emph{Algorithmica}, \textbf{85}(5):1544--1558).
[ { "created": "Thu, 17 Feb 2022 19:13:44 GMT", "version": "v1" } ]
2022-02-21
[ [ "Bose", "Prosenjit", "" ], [ "Morin", "Pat", "" ], [ "Odak", "Saeed", "" ] ]
The \emph{Product Structure Theorem} for planar graphs (Dujmovi\'c et al.\ \emph{JACM}, \textbf{67}(4):22) states that any planar graph is contained in the strong product of a planar $3$-tree, a path, and a $3$-cycle. We give a simple linear-time algorithm for finding this decomposition as well as several related decompositions. This improves on the previous $O(n\log n)$ time algorithm (Morin.\ \emph{Algorithmica}, \textbf{85}(5):1544--1558).
2105.01913
Samin Aref
Samin Aref and Zachary P. Neal
Identifying hidden coalitions in the US House of Representatives by optimally partitioning signed networks based on generalized balance
Post-peer-review version, 23 pages, 7 figures, 2 tables, combined article and supplementary information
null
10.1038/s41598-021-98139-w
null
cs.SI cs.CY math.OC physics.soc-ph
http://arxiv.org/licenses/nonexclusive-distrib/1.0/
In network science, identifying optimal partitions of a signed network into internally cohesive and mutually divisive clusters based on generalized balance theory is computationally challenging. We reformulate and generalize two binary linear programming models that tackle this challenge, demonstrating their practicality by applying them them to partition networks of collaboration in the US House of Representatives. These models guarantee a globally optimal network partition and can be practically applied to signed networks containing up to 30,000 edges. In the US House context, we find that a three-cluster partition is better than a conventional two-cluster partition, where the otherwise hidden third coalition is composed of highly effective legislators who are ideologically aligned with the majority party.
[ { "created": "Wed, 5 May 2021 07:57:41 GMT", "version": "v1" }, { "created": "Thu, 6 May 2021 05:17:10 GMT", "version": "v2" }, { "created": "Mon, 6 Sep 2021 22:17:22 GMT", "version": "v3" }, { "created": "Wed, 27 Jul 2022 00:54:13 GMT", "version": "v4" } ]
2022-07-28
[ [ "Aref", "Samin", "" ], [ "Neal", "Zachary P.", "" ] ]
In network science, identifying optimal partitions of a signed network into internally cohesive and mutually divisive clusters based on generalized balance theory is computationally challenging. We reformulate and generalize two binary linear programming models that tackle this challenge, demonstrating their practicality by applying them them to partition networks of collaboration in the US House of Representatives. These models guarantee a globally optimal network partition and can be practically applied to signed networks containing up to 30,000 edges. In the US House context, we find that a three-cluster partition is better than a conventional two-cluster partition, where the otherwise hidden third coalition is composed of highly effective legislators who are ideologically aligned with the majority party.