检索结果-内蒙古大学图书馆

arXiv 2017年

作者： Chaudhari, Pratik Baldassi, Carlo Zecchina, Riccardo Soatto, Stefano Talwalkar, Ameet Oberman, Adam Computer Science Department University of California Los Angeles Institute for Data Science and Analytics Bocconi University Milano Italy Department of Mathematics and Statistics McGill University Montreal Canada

We propose a new algorithm called Parle for parallel training of deep networks that converges 2-4× faster than a data-parallel implementation of SGD, while achieving significantly improved error rates that are nearly state-of-the-art on several benchmarks including CIFAR-10 and CIFAR-100, without introducing any additional hyper-parameters. We exploit the phenomenon of flat minima that has been shown to lead to improved generalization error for deep networks. Parle requires very infrequent communication with the parameter server and instead performs more computation on each client, which makes it well-suited to both single-machine, multi-GPU settings and distributed implementations. Copyright © 2017, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

NL-Augmenter A Framework for Task-Sensitive Natural Language Augmentation

arXiv

引用

arXiv 2021年

作者： Dhole, Kaustubh D. Gangal, Varun Gehrmann, Sebastian Gupta, Aadesh Li, Zhenhao Mahamood, Saad Mahendiran, Abinaya Mille, Simon Shrivastava, Ashish Tan, Samson Wu, Tongshuang Sohl-Dickstein, Jascha Choi, Jinho D. Hovy, Eduard Dušek, Ondřej Ruder, Sebastian Anand, Sajant Aneja, Nagender Banjade, Rabin Barthe, Lisa Behnke, Hanna Berlot-Attwell, Ian Boyle, Connor Brun, Caroline Sobrevilla Cabezudo, Marco Antonio Cahyawijaya, Samuel Chapuis, Emile Che, Wanxiang Choudhary, Mukund Clauss, Christian Colombo, Pierre Cornell, Filip Dagan, Gautier Das, Mayukh Dixit, Tanay Dopierre, Thomas Dray, Paul-Alexis Dubey, Suchitra Ekeinhor, Tatiana Giovanni, Marco Di Goyal, Tanya Gupta, Rishabh Hamla, Louanes Han, Sang Harel-Canada, Fabrice Honoré, Antoine Jindal, Ishan Joniak, Przemyslaw K. Kleyko, Denis Kovatchev, Venelin Krishna, Kalpesh Kumar, Ashutosh Langer, Stefan Lee, Seungjae Ryan Levinson, Corey James Liang, Hualou Liang, Kaizhao Liu, Zhexiong Lukyanenko, Andrey Marivate, Vukosi de Melo, Gerard Meoni, Simon Meyer, Maxime Mir, Afnan Moosavi, Nafise Sadat Muennighoff, Niklas Hon Mun, Timothy Sum Murray, Kenton Namysl, Marcin Obedkova, Maria Oli, Priti Pasricha, Nivranshu Pfister, Jan Plant, Richard Prabhu, Vinay Pais, Vasile Qin, Libo Raji, Shahab Rajpoot, Pawan Kumar Raunak, Vikas Rinberg, Roy Roberts, Nicholas Rodriguez, Juan Diego Roux, Claude Vasconcellos, P.H.S. Sai, Ananya B. Schmidt, Robin M. Scialom, Thomas Sefara, Tshephisho Shamsi, Saqib N. Shen, Xudong Shi, Yiwen Shi, Haoyue Shvets, Anna Siegel, Nick Sileo, Damien Simon, Jamie Singh, Chandan Sitelew, Roman Soni, Priyank Sorensen, Taylor Soto, William Srivastava, Aman Aditya Srivatsa, K.V. Sun, Tony Mukund Varma, T. Tabassum, A. Tan, Fiona Anting Teehan, Ryan Tiwari, Mo Tolkiehn, Marie Wang, Athena Wang, Zijian Wang, Zijie J. Wang, Gloria Wei, Fuxuan Wilie, Bryan Winata, Genta Indra Wu, Xinyi Wydmanski, Witold Xie, Tianbao Yaseen, Usama Yee, Michael A. Zhang, Jing Zhang, Yue ACKO Agara Amelia R&D New York United States Applied Research Laboratories The University of Texas at Austin United States Bloomberg Brigham Young University United States Carnegie Mellon University United States Center for Data and Computing in Natural Sciences Universität Hamburg Germany Charles River Analytics Charles University Prague Czech Republic Columbia University United States Council for Scientific and Industrial Research DeepMind United Kingdom Department of Computer Science University of Pretoria South Africa Drexel University United States Eberhard Karls University of Tübingen Germany Edinburgh Napier University United Kingdom Emory University United States Fablab by Inetum in Paris France Fraunhofer IAIS Germany Georgia Tech United States Google Brain United States Google Research United States Harbin Institute of Technology China Hasso Plattner Institute University of Potsdam Germany Hong Kong University of Science and Technology Hong Kong IBM Research IIT Delhi India IIT Madras India Illinois Mathematics and Science Academy United States Imperial College London United Kingdom Independent Indian Institute of Science Bangalore India Institut Teknologi Bandung Indonesia Institute of Data Science National University of Singapore Singapore International Institute of Information Technology Hyderabad India Jagiellonian University Poland Jean Monnet University France Johns Hopkins United States KTH Royal Institute of Technology Sweden KU Leuven Belgium MTS AI France Microsoft RedmondWA United States Mphasis NEXT Labs National University of Ireland Galway Ireland National University of Science and Technology Pakistan National University of Singapore Singapore Naver Labs Europe France Peking University China Politecnico di Milano University of Bologna Italy Polytechnic Institute of Paris France Pompeu Fabra University Spain Pontifical Catholic University of Minas Gerais Brazil Princeton University United States Rakuten India India Research Institu

data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Python-based natural language augmentation framework which supports the creation of both transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of natural language tasks. We demonstrate the efficacy of NL-Augmenter by using several of its tranformations to analyze the robustness of popular natural language models. The infrastructure, datacards and robutstness analysis results are available publicly on the NL-Augmenter repository (https://***/GEM-benchmark/NL-Augmenter). © 2021, CC BY.

关键词： Python

来源：评论

学校读者我要写书评

暂无评论

An improved metaheuristic algorithm for maximizing demand satisfaction in the population harvest cutting stock problem 9

An improved metaheuristic algorithm for maximizing demand sa...

引用

9th Annual Symposium on Combinatorial Search, SoCS 2016

作者： Climent, Laura O’Sullivan, Barry Wallace, Richard J Insight Centre for Data Analytics Department of Computer Science University College Cork Ireland

ISBN: (纸本)9781577357698

We present a greedy version of an existing metaheuristic algorithm for a special version of the Cutting Stock Problem (CSP). For this version, it is only possible to have indirect control over the patterns via a vector of continuous values which we refer to as a weights vector. Our algorithm iteratively generates new weights vectors by making local changes over the best weights vector computed so far. This allows us to achieve better solutions much faster than is possible with the original metaheuristic. Copyright © 2016, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Revisiting two-sided stability constraints

Revisiting two-sided stability constraints

引用

Douziemes Journees Francophones de Programmation par Contraintes, JFPC 2016 - 12th French-Speaking Conference on Constraint Programming, JFPC 2016

作者： Siala, Mohamed O'Sullivan, Barry Insight Centre for Data Analytics Department of Computer Science University College Cork Ireland

来源：评论

学校读者我要写书评

暂无评论

Causal discovery by randomness test

Causal discovery by randomness test

引用

2016 International Symposium on Artificial Intelligence and Mathematics, ISAIM 2016

作者： Prestwich, S. Tarim, S.A. Ozkan, I. Insight Centre for Data Analytics Department of Computer Science University College Cork Ireland

Probabilistic methods for causal discovery are based on the detection of patterns of correlation between variables. They are based on statistical theory and have revolutionised the study of causality. However, when correlation itself is unreliable, so are probabilistic methods: nonsense correlations can lead to spurious causal links, while nonmonotonic functional relationships between variables can prevent the detection of causal links. We describe a new heuristic method for inferring causality between two continuous or integer variables, based on a nonparametric randomness test. We evaluate the accuracy of the method by comparing it to published algorithms on real and artificial datasets, and show that it largely avoids these false positives and negatives. © 2016 University of Virginia. All rights reserved.

关键词： Heuristic methods

来源：评论

学校读者我要写书评

暂无评论

NeoN: Neuromorphic control for autonomous robotic navigation

NeoN: Neuromorphic control for autonomous robotic navigation

引用

IEEE International Symposium on Robotics and Intelligent Sensors (IRIS)

作者： J. Parker Mitchell Grant Bruer Mark E. Dean James S. Plank Garrett S. Rose Catherine D. Schuman Department of Electrical Engineering and Computer Science University of Tennessee Knoxville TN Computational Data Analytics Oak Ridge National Laboratory Oak Ridge TN

In this paper we describe the use of a new neuromorphic computing framework to implement the navigation system for a roaming, obstacle avoidance robot. Using a Dynamic Adaptive Neural Network Array (DANNA) structure, our TENNLab (Laboratory of Tennesseans Exploring Neural Networks) hardware/software co-design framework and evolutionary optimization (EO) as the training algorithm, we create, train, implement, and test a spiking neural network autonomous robot control system using an array of neuromorphic computing elements built on an FPGA. The simplicity and flexibility of the DANNA neuromorphic computing elements allow for sufficient scale and connectivity on a Xilinx Kintex-7 FPGA to support sensory input and motor control for a mobile robot to navigate a dynamically changing environment. We further describe how more complex capabilities can be added using the same platform, e.g. object identification and tracking.

关键词： Neurons Robot sensing systems Neuromorphics Biological neural networks computer architecture

来源：评论

学校读者我要写书评

暂无评论

Modified cuckoo search ased neural networks for forest types classification 2

Modified cuckoo search ased neural networks for forest types...

引用

2nd International Conference on Information Technology and Intelligent Transportation Systems, ITITS 2017

作者： Chatterjee, Sankhadeep Dey, Nilanjan Sen, Soumya Ashour, Amira S. Fong, Simon James Shi, Fuqian Department of Computer Science and Engineering University of Calcutta Kolkata India Department of Information Technology Techno India College of Technology Kolkata India A. K. Choudhury School of Information Technology University of Calcutta Kolkata West Bengal India Department of Electronics and Electrical Communications Engineering Faculty of Engineering Tanta University Egypt Department of Computer and Information Science Data Analytics and Collaborative Computing Laboratory University of Macau Taipa China College of Information and Engineering Wenzhou Medical University Wenzhou China

ISBN: (纸本)9781614997849

Pixel classification in land scape images is a challenging process especially in forest images due to the similar spectral features of pixels situated close to each other. Previously, meta-heuristic coupled artificial neural network (ANN) models have been used to classify the two-different species, namely Japanese Cedar, Japanese Cypress and one mixed forest class. Previous attempts have shown reasonable improvement in the classification process using genetic algorithm (GA) supported neural network over other traditional approaches. Consequently, in the current work, a modified Cuckoo Search (CS) supported Neural Network (NN-MCS) classifier is proposed. The lévy flight associated with cuckoo search has been modified using McCulloch's method of generating stable random numbers. The proposed approach is compared with GA-NN using single objective function and CS-NN (ANN trained with CS) classifiers in terms of confusion matrix based performance metrics. The results depicted the dominance of the suggested NN-MCS model compared to the CS-NN model to a greater extent. © 2017 The authors and IOS Press. All rights reserved.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

Preprocessing versus search processing for constraint satisfaction problems 23

Preprocessing versus search processing for constraint satisf...

引用

23rd RCRA International Workshop on Experimental Evaluation of Algorithms for Solving Problems with Combinatorial Explosion, RCRA 2016

作者： Wallace, Richard J. Insight Centre for Data Analytics Department of Computer Science University College Cork Cork Ireland

A perennial problem in hybrid backtrack CSP search is how much local consistency processing should be done to achieve the best efficiency. This can be divided into two separate questions: (1) how much work should be done before the actual search begins, i.e. during preprocessing?, and (2) how much of the same processing should be interleaved with search? At present there are two leading approaches to establishing stronger consistencies than the basic arc consistency maintenance that is done in most solvers. On the one hand there are various kinds singleton arc consistency that can be used;on the other there are several variants of restricted path consistency. To date these have not been compared directly. The present work attempts to do this for a variety of problems, and in so doing, it also provides an empirical evaluation of the preprocessing versus search processing issue. Comparisons are made using the domain/degree and domain/weighted degree variable ordering heuristics. In general, it appears that preprocessing with higher levels of consistency followed by hybrid-AC processing (i.e. MAC) gives the best results, especially when the weighted degree heuristic is used. For problems with n-ary constraints, this difference seems to be even more pronounced. In some cases, higher levels of consistency maintenance established during preprocessing leads to performance gains over MAC of several orders of magnitude.

关键词： Constraint satisfaction problems

来源：评论

学校读者我要写书评

暂无评论

LEADER FORMATION WITH MEAN-FIELD BIRTH AND DEATH MODELS

arXiv

引用

arXiv 2018年

作者： Albi, Giacomo Bongini, Mattia Rossi, Francesco Solombrino, Francesco Department of Computer Science University of Verona Str. Le Grazie 15 VeronaIT-37134 Italy Big Data and Marketing Analytics CRIF Via M. Fantin 3 BolognaIT-40131 Italy Department of Mathematics "Tullio Levi-Civita" University of Padova Via Trieste 63 PadovaIT-35121 Italy Department of Mathematics and Applications "R. Caccioppoli" University of Naples "Federico II" Via Cintia Monte S. Angelo NaplesIT-80126 Italy

We provide a mean-field description for a leader-follower dynamics with mass transfer among the two populations. This model allows the transition from followers to leaders and vice versa, with scalar-valued transition rates depending nonlinearly on the global state of the system at each time. We first prove the existence and uniqueness of solutions for the leader-follower dynamics, under suitable assumptions. We then establish, for an appropriate choice of the initial datum, the equivalence of the system with a PDE-ODE system, that consists of a continuity equation over the state space and an ODE for the transition from leader to follower or vice versa. We further introduce a stochastic process approximating the PDE, together with a jump process that models the switch between the two populations. Using a propagation of chaos argument, we show that the particle system generated by these two processes converges in probability to a solution of the PDE-ODE system. Finally, several numerical simulations of social interactions dynamics modeled by our system are *** Codes 35Q82, 35Q91, 70F45, 65M08, 82C22, 82B21 Copyright © 2018, The Authors. All rights reserved.

关键词： Markov processes

来源：评论

学校读者我要写书评

暂无评论

Neighbourhood SAC for constraint satisfaction problems with non-binary constraints 29

Neighbourhood SAC for constraint satisfaction problems with ...

引用

29th International Florida Artificial Intelligence Research Society Conference, FLAIRS 2016

作者： Wallace, Richard J. Insight Centre for Data Analytics Department of Computer Science University College Cork Cork Ireland

ISBN: (纸本)9781577357568

Neighbourhood singleton arc consistency (NSAC) is a type of singleton arc consistency (SAC) in which the subproblem formed by variables adjacent to a variable with a singleton domain is made arc consistent. In this paper we consider how to apply this form of consistency reasoning to problems with n-ary constraints including global constraints. The chief problem encountered is that of neighbouring variables contained in a constraint that also includes non-neighbouring variables. In this case, a strict extension of NSAC involves projection of such constraints onto the neighbourhood variables, but for many global constraints this may be difficult to do in practice. Here, we consider a simple variant called restricted neighbourhood SAC, that avoids this problem. We compare the two approaches on random and structured problems and show that in all cases the restricted form of fc-NSAC is nearly as effective as the unrestricted form. © 2016, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Constraint satisfaction problems

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：