检索结果-内蒙古大学图书馆

Top-rank enhanced listwise optimization for statistical machine translation

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Chen, Huadong Huang, Shujian Chiang, David Dai, Xinyu Chen, Jiajun State Key Laboratory for Novel Software Technology Nanjing University Department of Computer Science and Engineering University of Notre Dame

Pairwise ranking methods are the basis of many widely used discriminative training approaches for structure prediction problems in natural language processing (NLP). Decomposing the problem of ranking hypotheses into pairwise comparisons enables simple and efficient solutions. However, neglecting the global ordering of the hypothesis list may hinder learning. We propose a listwise learning framework for structure prediction problems such as machine translation. Our framework directly models the entire translation list's ordering to learn parameters which may better fit the given listwise samples. Furthermore, we propose top-rank enhanced loss functions, which are more sensitive to ranking errors at higher positions. Experiments on a large-scale Chinese-English translation task show that both our listwise learning framework and top-rank enhanced listwise losses lead to significant improvements in translation quality. Copyright © 2017, The Authors. All rights reserved.

关键词： computer aided language translation

Deep discrete supervised hashing

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Jiang, Qing-Yuan Cui, Xue Li, Wu-Jun National Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing University Nanjing210023 China

Hashing has been widely used for large-scale search due to its low storage cost and fast query speed. By using supervised information, supervised hashing can significantly outperform unsupervised hashing. Recently, discrete supervised hashing and deep hashing are two representative progresses in supervised hashing. On one hand, hashing is essentially a discrete optimization problem. Hence, utilizing supervised information to directly guide discrete (binary) coding procedure can avoid sub-optimal solution and improve the accuracy. On the other hand, deep hashing, which integrates deep feature learning and hash-code learning into an end-to-end architecture, can enhance the feedback between feature learning and hash-code learning. The key in discrete supervised hashing is to adopt supervised information to directly guide the discrete coding procedure in hashing. The key in deep hashing is to adopt the supervised information to directly guide the deep feature learning procedure. However, there have not existed works which can use the supervised information to directly guide both discrete coding procedure and deep feature learning procedure in the same framework. In this paper, we propose a novel deep hashing method, called deep discrete supervised hashing (DDSH), to address this problem. DDSH is the first deep hashing method which can utilize supervised information to directly guide both discrete coding procedure and deep feature learning procedure, and thus enhance the feedback between these two important procedures. Experiments on three real datasets show that DDSH can outperform other state-of-the-art baselines, including both discrete hashing and deep hashing baselines, for image retrieval. Copyright © 2017, The Authors. All rights reserved.

关键词： Hash functions

A B-spline curve extension algorithm

学校读者我要写书评

暂无评论

science China(Information sciences) 2016年第3期59卷 37-45页

作者： Yang LU Kanle SHI Junhai YONG Hejin GU Haichuan SONG School of Software Tsinghua University Department of Computer Science and Technology Tsinghua University Key Laboratory for Information System Security Ministry of Education of China Tsinghua National Laboratory for Information Science and Technology Jiangxi Academy of Science

B-spline curve extension is an important operation in computer aided design systems. In this paper,we present a new extension algorithm for B-spline curves. The algorithm uses curve unclamping to generate a uniform B-spline curve segment from the original curve and gradually extends the segment to pass through every target point. Algorithms of uniform B-spline curves are used such that our algorithm has a low time cost and can easily handle arbitrary-order derivative constraints at the target points. Generalization for non-uniform rational B-spline curve extension is also discussed, and examples show the efficiency of our method.

关键词： curve extension B-spline/NURBS unclamping clamping uniform

Discrete latent factor model for cross-modal hashing

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Jiang, Qing-Yuan Li, Wu-Jun National Key Laboratory for Novel Software Technology Collaborative Innovation Center of Novel Software Technology and Industrialization Department of Computer Science and Technology Nanjing University China

Due to its storage and retrieval efficiency, cross-modal hashing (CMH) has been widely used for cross-modal similarity search in multimedia applications. According to the training strategy, existing CMH methods can be mainly divided into two categories: relaxation-based continuous methods and discrete methods. In general, the training of relaxation-based continuous methods is faster than discrete methods, but the accuracy of relaxation-based continuous methods is not satisfactory. On the contrary, the accuracy of discrete methods is typically better than relaxation-based continuous methods, but the training of discrete methods is time-consuming. In this paper, we propose a novel CMH method, called discrete latent factor model based cross-modal hashing (DLFH), for cross modal similarity search. DLFH is a discrete method which can directly learn the binary hash codes for CMH. At the same time, the training of DLFH is efficient. Experiments on real datasets show that DLFH can achieve significantly better accuracy than existing methods, and the training time of DLFH is comparable to that of relaxation-based continuous methods which are much faster than existing discrete methods. Copyright © 2017, The Authors. All rights reserved.

关键词： Hash functions

Asymmetric deep supervised hashing

学校读者我要写书评

暂无评论

arXiv 2017年

Hashing has been widely used for large-scale approximate nearest neighbor search because of its storage and search efficiency. Recent work has found that deep supervised hashing can significantly outperform non-deep supervised hashing in many applications. However, most existing deep supervised hashing methods adopt a symmetric strategy to learn one deep hash function for both query points and database (retrieval) points. The training of these symmetric deep supervised hashing methods is typically time-consuming, which makes them hard to effectively utilize the supervised information for cases with large-scale database. In this paper, we propose a novel deep supervised hashing method, called asymmetric deep supervised hashing (ADSH), for large-scale nearest neighbor search. ADSH treats the query points and database points in an asymmetric way. More specifically, ADSH learns a deep hash function only for query points, while the hash codes for database points are directly learned. The training of ADSH is much more efficient than that of traditional symmetric deep supervised hashing methods. Experiments show that ADSH can achieve state-of-the-art performance in real applications. Copyright © 2017, The Authors. All rights reserved.

关键词： Nearest neighbor search

Joint product numerical range and geometry of reduced density matrices

学校读者我要写书评

暂无评论

science China(Physics,Mechanics & Astronomy) 2017年第2期60卷 9-17页

作者： Jianxin Chen Cheng Guo Zhengfeng Ji Yiu-Tung Poon Nengkun YU Bei Zeng Jie Zhou Joint Center for Quantum Information and Computer Science University of Maryland College Park 20740 USA Institute for Advanced Study Tsinghua University Beijing 100084 China Centre for Quantum Computation ＆ Intelligent Systems School of Software Faculty of Engineering and Information Technology University of Technology Sydney Sydney 2007 Australia State Key Laboratory of Computer Science.Institute of Software Chinese Academy of Sciences Beijing 100084 China Department of Mathematics Iowa State University Ames 50011-2140 USA Institute for Quantum Computing University of Waterloo Waterloo N2L 3G1 Canada Department of Mathematics ＆ Statistics University of Guelph Guelph N1G 2 W1 Canada Perimeter Institute for Theoretical Physics Waterloo N2L 2Y5 Canada

The reduced density matrices of a many-body quantum system form a convex set, whose three-dimensional projection is convex in R3. The boundary of may exhibit nontrivial geometry, in particular ruled surfaces. Two physical mechanisms are known for the origins of ruled surfaces： symmetry breaking and gapless. In this work, we study the emergence of ruled surfaces for systems with local Hamiltonians in infinite spatial dimension, where the reduced density matrices are known to be separable as a consequence of the quantum de Finetti＇s theorem. This allows us to identify the reduced density matrix geometry with joint product numerical range II of the Hamiltonian interaction terms. We focus on the case where the interaction terms have certain structures, such that a ruled surface emerges naturally when taking a convex hull of ∏. We show that, a ruled surface on sitting in ∏ has a gapless origin, otherwise it has a symmetry breaking origin. As an example, we demonstrate that a famous ruled surface, known as the oloid, is a possible shape of , with two boundary pieces of symmetry breaking origin separated by two gapless lines.

关键词： reduced density matrices quantum tomography numerical range quantum information oloid ruled surface

A Simple Analysis for Exp-concave Empirical Minimization with Arbitrary Convex Regularizer

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Yang, Tianbao Li, Zhe Zhang, Lijun Department of Computer Science University of Iowa National Key Laboratory for Novel Software Technology Nanjing University

In this paper, we present a simple analysis of fast rates with high probability of empirical minimization for stochastic composite optimization over a finite-dimensional bounded convex set with exponential concave loss functions and an arbitrary convex regularization. To the best of our knowledge, this result is the first of its kind. As a byproduct, we can directly obtain the fast rate with high probability for exponential concave empirical risk minimization with and without any convex regularization, which not only extends existing results of empirical risk minimization but also provides a unified framework for analyzing exponential concave empirical risk minimization with and without any convex regularization. Our proof is very simple only exploiting the covering number of a finite-dimensional bounded set and a concentration inequality of random vectors. Copyright © 2017, The Authors. All rights reserved.

关键词： Stochastic systems

A Comparison of CPU and GPU implementations for the LHCb Experiment Run 3 Trigger

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Aaij, R. Adinolfi, M. Aiola, S. Akar, S. Albrecht, J. Alexander, M. Amato, S. Amhis, Y. Archilli, F. Bala, M. Bassi, G. Bian, L. Blago, M.P. Boettcher, T. Boldyrev, A. Borghi, S. Brea Rodriguez, A. Calefice, L. Calvo Gomez, M. Cámpora Pérez, D.H. Cardini, A. Cattaneo, M. Chobanova, V. Ciezarek, G. Cid Vidal, X. Cobbledick, J.L. Coelho, J.A.B. Colombo, T. Contu, A. Couturier, B. Craik, D.C. Currie, R. d’Argent, P. de Cian, M. Derkach, D. Dordei, F. Dorigo, M. Dufour, L. Durante, P. Dziurda, A. Dzyuba, A. Easo, S. Esen, S. Fernandez Declara, P. Filippov, S. Fitzpatrick, C. Frank, M. Gandini, P. Gligorov, Vladimir Vava Golobardes, E. Graziani, G. Grillo, L. Günther, P.A. Hansmann-Menzemer, S. Hennequin, A.M. Henry, L. Hill, D. Hollitt, S.E. Hu, J. Hulsbergen, W. Hunter, R.J. Hushchyn, M. Jashal, B.K. Jones, C.R. Klaver, S. Klimaszewski, K. Kopecna, R. Krzemien, W. Kucharczyk, M. Lane, R. Lazzari, F. Le Gac, R. Li, P. Lopes, J.H. Lucio Martinez, M. Lupato, A. Lupton, O. Lyu, X. Machefert, F. Madejczyk, O. Malde, S. Marchand, J.F. Mariani, S. Marin Benito, C. Martinez Santos, D. Martinez Vidal, F. Matev, R. Mazurek, M. Mitreska, B. Mitzel, D.S. Morello, M.J. Mu, H. Muzzetto, P. Naik, P. Needham, M. Neri, N. Neufeld, N. Nolte, N.S. O’Hanlon, D. Oyanguren, A. Pepe Altarelli, M. Petrucci, S. Petruzzo, M. Pica, L. Pisani, F. Piucci, A. Polci, F. Poluektov, A. Polycarpo, E. Prouve, C. Punzi, G. Quagliani, R. Rabadan Trejo, R.I. Ramos Pernas, M. Rangel, M.S. Ratnikov, F. Raven, G. Reiss, F. Renaudin, V. Robbe, P. Ryzhikov, A. Santimaria, M. Saur, M. Schiller, M. Schwemmer, R. Sciascia, B. Solomin, A. Suljik, F. Skidmore, N. Sokoloff, M.D. Spradlin, P. Stahl, M. Stahl, S. Stevens, H. Sun, L. Szabelski, A. Szumlak, T. Szymanski, M. Tou, D.Y. Tuci, G. Usachov, A. Valls Canudas, N. Vazquez Gomez, R. Vecchi, S. Vesterinen, M. Vilasis-Cardona, X. Vom Bruch, D. Wang, Z. Wojton, T. Whitehead, M. Williams, M. Witek, M. Xie, Y. Xu, A. Yin, H. Zdybal, M. Zenaiev, O. Zhang, D. Zhang, L. Zhu, X. Rio de Janeiro Brazil Center for High Energy Physics Tsinghua University Beijing China School of Physics State Key Laboratory of Nuclear Physics and Technology Peking University Beijing China University of Chinese Academy of Sciences Beijing China Institute of Particle Physics Central China Normal University Hubei Wuhan China Univ. Grenoble Alpes Univ. Savoie Mont Blanc CNRS IN2P3-LAPP Annecy France Aix Marseille Univ CNRS IN2P3 CPPM Marseille France Université Paris-Saclay CNRS IN2P3 IJCLab Orsay France LPNHE Sorbonne Université Paris Diderot Sorbonne Paris Cité CNRS IN2P3 Paris France Fakultät Physik Technische Universität Dortmund Dortmund Germany Physikalisches Institut Ruprecht-Karls-Universität Heidelberg Heidelberg Germany INFN Sezione di Ferrara Ferrara Italy INFN Sezione di Firenze Firenze Italy INFN Laboratori Nazionali di Frascati Frascati Italy INFN Sezione di Milano Milano Italy INFN Sezione di Cagliari Monserrato Italy INFN Sezione di Pisa Pisa Italy Nikhef National Institute for Subatomic Physics Amsterdam Netherlands Nikhef National Institute for Subatomic Physics VU University Amsterdam Amsterdam Netherlands AGH - University of Science and Technology Faculty of Physics and Applied Computer Science Kraków Poland Henryk Niewodniczanski Institute of Nuclear Physics Polish Academy of Sciences Kraków Poland Warsaw Poland Gatchina Russia Moscow Russia ICCUB Universitat de Barcelona Barcelona Spain Universidade de Santiago de Compostela Santiago de Compostela Spain Instituto de Fisica Corpuscular Centro Mixto Universidad de Valencia - CSIC Valencia Spain Geneva Switzerland Lausanne Switzerland Physik-Institut Universität Zürich Zürich Switzerland H.H. Wills Physics Laboratory University of Bristol Bristol United Kingdom Cavendish Laboratory University of Cambridge Cambridge United Kingdom Department of Physics University of Warwick Coventry United Kingdom STFC Rutherford Appleton Laboratory Didcot Uni

The LHCb experiment at CERN is undergoing an upgrade in preparation for the Run 3 data taking period of the LHC. As part of this upgrade the trigger is moving to a fully software implementation operating at the LHC bunch crossing rate. We present an evaluation of a CPU-based and a GPU-based implementation of the first stage of the High Level Trigger. After a detailed comparison both options are found to be viable. This document summarizes the performance and implementation details of these options, the outcome of which has led to the choice of the GPU-based implementation as the baseline. © 2021, CC BY.

关键词： Hadrons

学校读者我要写书评

暂无评论

基于驾驶脑的智能驾驶车辆硬件平台架构

Engineering 2018年第4期4卷 78-92页

作者：李德毅高洪波 Department of Computer Science and Technology Tsinghua University State Key Laboratory of Automotive Safety and Energy Tsinghua University Center for Intelligent Connected Vehicles and Transportation Tsinghua University

不同的智能驾驶试验平台的传感器型号、数量、安装位置各不相同,导致传感器信息处理模块也各不相同;不同的驾驶地图,其提供信息的粒度也没有固定标准,由此构成的智能驾驶系统软件模块的数量、接口各不相同。基于以驾驶脑为核心的智能驾... 详细信息

不同的智能驾驶试验平台的传感器型号、数量、安装位置各不相同,导致传感器信息处理模块也各不相同;不同的驾驶地图,其提供信息的粒度也没有固定标准,由此构成的智能驾驶系统软件模块的数量、接口各不相同。基于以驾驶脑为核心的智能驾驶车辆软件与硬件架构,决策模块将不直接与传感器信息处理模块发生关联,通过驾驶认知的形式化语言,将驾驶认知形式化,由驾驶脑认知形成决策。驾驶认知的形式化降低了传感器数量、类型、安装位置的变化对整个软件架构的影响,使得软件架构可以在不同传感器配置车辆平台上方便地移植。

关键词：驾驶脑智能驾驶硬件平台架构