检索结果-内蒙古大学图书馆

The Minority Matters: A Diversity-Promoting Collaborative Metric Learning Algorithm

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Bao, Shilong Xu, Qianqian Yang, Zhiyong He, Yuan Cao, Xiaochun Huang, Qingming State Key Laboratory of Information Security Institute of Information Engineering CAS China School of Cyber Security University of Chinese Academy of Sciences China Key Lab. of Intelligent Information Processing Institute of Computing Technology CAS China School of Computer Science and Tech. University of Chinese Academy of Sciences China Alibaba Group China School of Cyber Science and Technology Shenzhen Campus Sun Yat-sen University China Key Laboratory of Big Data Mining and Knowledge Management CAS China Peng Cheng Laboratory China

Collaborative Metric Learning (CML) has recently emerged as a popular method in recommendation systems (RS), closing the gap between metric learning and Collaborative Filtering. Following the convention of RS, existing methods exploit unique user representation in their model design. This paper focuses on a challenging scenario where a user has multiple categories of interests. Under this setting, we argue that the unique user representation might induce preference bias, especially when the item category distribution is imbalanced. To address this issue, we propose a novel method called Diversity-Promoting Collaborative Metric Learning (DPCML), with the hope of considering the commonly ignored minority interest of the user. The key idea behind DPCML is to include a multiple set of representations for each user in the system. Based on this embedding paradigm, user preference toward an item is aggregated from different embeddings by taking the minimum item-user distance among the user embedding set. Furthermore, we observe that the diversity of the embeddings for the same user also plays an essential role in the model. To this end, we propose a diversity control regularization term to accommodate the multi-vector representation strategy better. Theoretically, we show that DPCML could generalize well to unseen test data by tackling the challenge of the annoying operation that comes from the minimum value. Experiments over a range of benchmark datasets speak to the efficacy of DPCML. © 2022, CC BY-NC-SA.

关键词： Embeddings

The Adaptive Fault-tolerant Routing Based on an Improved Local Security Information Model of the Exchanged Hypercube

学校读者我要写书评

暂无评论

The Adaptive Fault-tolerant Routing Based on an Improved Loc...

IEEE International Conference on big data and Cloud computing (BdCloud)

作者： Yazhi Zhang Chuanwen Luo Guijuan Wang Li Zhang Mengjie Lv Jiguo Yu School of Computer Science and Technology Qilu University of Technology Jinan China School of Information Science and Technology Beijing Forestry University Beijing China Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center Qilu University of Technology (Shandong Academy of Sciences) Jinan China Shandong Provincial Key Laboratory of Computer Networks Shandong Fundamental Research Center for Computer Science Jinan China College of Computer Nanjing University of Posts and Telecommunications Nanjing China Big Data Research Institute Qilu University of Technology (Shandong Academy of Sciences) Jinan China

With the increasing amount of computation in high-performance computing, the scale of interconnection networks is becoming larger and larger. It is inevitable that processors or links in the network become faulty. Therefore, it is very important for a processor to understand the security status of itself and neighbors, and then adaptively and reliably communicate. However, the existing security information models consider faulty processors or faulty links separately, and do not consider the situation where both processors and links fail simultaneously. The Exchanged Hypercube is an excellent interconnection network structure which achieves a balance between network functionality and hardware overhead while preserving many excellent properties of the Hypercube. In this paper, we propose an adaptive fault-tolerant routing algorithm of the Exchanged Hypercube based on an improved local security information model, which considers the faulty processors and faulty links simultaneously. This algorithm always can select suitable processors for routing based on the local security information and unsafe coefficients of processors, which greatly reduces the impact of the faulty of networks. The adaptive fault-tolerant routing algorithm of the Exchanged Hypercube network is not only suitable for regular networks but also for heterogeneous networks since the Exchanged Hypercube is a semi regular network. Sufficient experimental results show that the algorithm proposed in this paper has a lower packet loss rate, a higher success rate, and a shorter average path length.

关键词：

Joint Compression and Deadline Optimization for Wireless Federated Learning

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhang, Maojun Li, Yang Liu, Dongzhu Jin, Richeng Zhu, Guangxu Zhong, Caijun Quek, Tony Q.S. The College of information Science and Electronic Engineering Zhejiang University Hangzhou China China Academy of Information and Communications Technology Beijing China School of Computing Science University of Glasgow United Kingdom The Department of Information and Communication Engineering Zhejiang University Hangzhou310007 China The Zhejiang–Singapore Innovation and AI Joint Research Lab Hangzhou310007 China Hangzhou310007 China The Shenzhen Research Institute of Big Data Guangdong Shenzhen518172 China The Peng Cheng Laboratory Guangdong Shenzhen518055 China Guangdong Guangzhou510555 China The Singapore University of Technology and Design Singapore487372 Singapore

Federated edge learning (FEEL) is a popular distributed learning framework for privacy-preserving at the edge, in which densely distributed edge devices periodically exchange model-updates with the server to complete the global model training. Due to limited bandwidth and uncertain wireless environment, FEEL may impose heavy burden to the current communication system. In addition, under the common FEEL framework, the server needs to wait for the slowest device to complete the update uploading before starting the aggregation process, leading to the straggler issue that causes prolonged communication time. In this paper, we propose to accelerate FEEL from two aspects: i.e., 1) performing data compression on the edge devices and 2) setting a deadline on the edge server to exclude the straggler devices. However, undesired gradient compression errors and transmission outage are introduced by the aforementioned operations respectively, affecting the convergence of FEEL as well. In view of these practical issues, we formulate a training time minimization problem, with the compression ratio and deadline to be optimized. To this end, an asymptotically unbiased aggregation scheme is first proposed to ensure zero optimality gap after convergence, and the impact of compression error and transmission outage on the overall training time are quantified through convergence analysis. Then, the formulated problem is solved in an alternating manner, based on which, the novel joint compression and deadline optimization (JCDO) algorithm is derived. Numerical experiments for different use cases in FEEL including image classification and autonomous driving show that the proposed method is nearly 30X faster than the vanilla FedSGD algorithm, and outperforms the state-of-the-art schemes. © 2023, CC BY.

关键词： Autonomous vehicles

AdAUC: End-to-end Adversarial AUC Optimization Against Long-tail Problems

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Hou, Wenzheng Xu, Qianqian Yang, Zhiyong Bao, Shilong He, Yuan Huang, Qingming Key Laboratory of Intelligent Information Processing Institute of Computing Technology CAS Beijing China School of Computer Science and Technology University of Chinese Academy of Sciences Beijing China State Key Laboratory of Information Security Institute of Information Engineering CAS Beijing China School of Cyber Security University of Chinese Academy of Sciences Beijing China Alibaba Group Beijing China Key Laboratory of Big Data Mining and Knowledge Management Chinese Academy of Sciences Beijing China Artificial Intelligence Research Center Peng Cheng Laboratory Shenzhen China

It is well-known that deep learning models are vulnerable to adversarial examples. Existing studies of adversarial training have made great progress against this challenge. As a typical trait, they often assume that the class distribution is overall balanced. However, long-tail datasets are ubiquitous in a wide spectrum of applications, where the amount of head class instances is larger than the tail classes. Under such a scenario, AUC is a much more reasonable metric than accuracy since it is insensitive toward class distribution. Motivated by this, we present an early trial to explore adversarial training methods to optimize AUC. The main challenge lies in that the positive and negative examples are tightly coupled in the objective function. As a direct result, one cannot generate adversarial examples without a full scan of the dataset. To address this issue, based on a concavity regularization scheme, we reformulate the AUC optimization problem as a saddle point problem, where the objective becomes an instance-wise function. This leads to an end-to-end training protocol. Furthermore, we provide a convergence guarantee of the proposed algorithm. Our analysis differs from the existing studies since the algorithm is asked to generate adversarial examples by calculating the gradient of a min-max problem. Finally, the extensive experimental results show the performance and robustness of our algorithm in three long-tail datasets. © 2022, CC BY.

关键词： Deep learning

Z-Inspection®: A Process to Assess Trustworthy AI

学校读者我要写书评

暂无评论

IEEE Transactions on Technology and Society

IEEE Transactions on Technology and Society 2021年第2期2卷 83-97页

作者： Zicari, Roberto V. Brodersen, John Brusseau, James Dudder, Boris Eichhorn, Timo Ivanov, Todor Kararigas, Georgios Kringen, Pedro McCullough, Melissa Moslein, Florian Mushtaq, Naveed Roig, Gemma Sturtz, Norman Tolle, Karsten Tithi, Jesmin Jahan Van Halem, Irmhild Westerlund, Magnus Frankfurt Big Data Lab Goethe University Frankfurt Frankfurt Germany Department of Public Health Section of General Practice and Research Unit for General Practice Faculty of Health and Medical Sciences University of Copenhagen Copenhagen Denmark Philosophy Department Pace University New YorkNY United States Department of Computer Science University of Copenhagen Copenhagen Denmark Department of Physiology Faculty of Medicine University of Iceland Iceland Institute of the Law and Regulation of Digitalization Philipps-University Marburg Marburg Germany Parallel Computing Labs Intel Labs Santa ClaraCA United States Department of Computer Science Arcada University of Applied Sciences Helsinki Finland Primary Health Care Research Unit Copenhagen1014 Denmark

The ethical and societal implications of artificial intelligence systems raise concerns. In this article, we outline a novel process based on applied ethics, namely, Z-Inspection®, to assess if an AI system is trustworthy. We use the definition of trustworthy AI given by the high-level European Commission's expert group on AI. Z-Inspection® is a general inspection process that can be applied to a variety of domains where AI systems are used, such as business, healthcare, and public sector, among many others. To the best of our knowledge, Z-Inspection® is the first process to assess trustworthy AI in practice. © 2020 IEEE.

关键词： Deep learning

Probabilistic methods for approximate archetypal analysis

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Han, Ruijian Osting, Braxton Wang, Dong Xu, Yiming Department of Statistics The Chinese University of Hong Kong Hong Kong Department of Mathematics University of Utah Salt Lake City United States School of Science and Engineering The Chinese University of Hong Kong Shenzhen China Guangdong Provincial Key Laboratory of Big Data Computing The Chinese University of Hong Kong Shenzhen China Scientific Computing and Imaging Institute University of Utah Salt Lake City United States

Archetypal analysis is an unsupervised learning method for exploratory data analysis. One major challenge that limits the applicability of archetypal analysis in practice is the inherent computational complexity of the existing algorithms. In this paper, we provide a novel approximation approach to partially address this issue. Utilizing probabilistic ideas from high-dimensional geometry, we introduce two preprocessing techniques to reduce the dimension and representation cardinality of the data, respectively. We prove that provided the data is approximately embedded in a low-dimensional linear subspace and the convex hull of the corresponding representations is well approximated by a polytope with a few vertices, our method can effectively reduce the scaling of archetypal analysis. Moreover, the solution of the reduced problem is near-optimal in terms of prediction errors. Our approach can be combined with other acceleration techniques to further mitigate the intrinsic complexity of archetypal analysis. We demonstrate the usefulness of our results by applying our method to summarize several moderately large-scale datasets. Copyright © 2021, The Authors. All rights reserved.

关键词： Large dataset

Algebraic Analysis of Bifurcations and Chaos for Discrete Dynamical Systems 8th

学校读者我要写书评

暂无评论

Algebraic Analysis of Bifurcations and Chaos for Discrete Dy...

8th International Conference on Mathematical Aspects of Computer and Information Sciences, MACIS 2019

作者： Huang, Bo Niu, Wei LMIB-School of Mathematical Sciences Beihang University Beijing100191 China Courant Institute of Mathematical Sciences New York University New York10012 United States Ecole Centrale de Pékin Beihang University Beijing100191 China Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University Beijing100191 China

ISBN: (纸本)9783030431198

This paper deals with the stability, bifurcations and chaotic behaviors of discrete dynamical systems by using methods of symbolic computation. We explain how to reduce the problems of analyzing the stability, bifurcations and chaos induced by snapback repellers to algebraic problems, and solve them by using an algorithmic approach based on methods for solving semi-algebraic systems. The feasibility of the symbolic approach is demonstrated by analyses of the dynamical behaviors for several discrete models. © 2020, Springer Nature Switzerland AG.

关键词： Algebra

Integral Sliding-Mode Control of Piecewise Linear systems

学校读者我要写书评

暂无评论

Integral Sliding-Mode Control of Piecewise Linear systems

第三十九届中国控制会议

作者： Chunyang Zhang Qing Gao Peng Zhang Kang Zhou School of Automation Science and Eectrical Engineering Beihang University Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University School of Mechatronical Engineering Beijing Institute of Technology

This paper proposes a dynamic integral sliding-mode control approach for a class of nonlinear systems based on piecewise affine linear models. A general nonlinear system is first modeled equivalently as a uncertain piecewise affine linear system with norm-bounded approximation errors. Based on the equivalent piecewise affine linear system, a dynamic integral sliding-mode control（DISMC） scheme is then proposed. The scheme holds a favorable characteristic in the sense that it allows different input matrices in each subregion, which makes the proposed approach more relaxed than many existing ***, a constructive procedure of the DISMC via convex optimization is presented. Finally, a numerical example is provided to demonstrate the advantages and effectiveness of the developed approach.

关键词： Dynamic integral sliding mode control piecewise affine linear models linear matrix inequality（LMI）

Neural posterior estimation with guaranteed exact coverage: The ringdown of GW150914

学校读者我要写书评

暂无评论

Physical Review D 2023年第4期108卷 044029-044029页

作者： Marco Crisostomi Kallol Dey Enrico Barausse Roberto Trotta SISSA Via Bonomea 265 34136 Trieste Italy and INFN Sezione di Trieste 34127 Trieste Italy IFPU—Institute for Fundamental Physics of the Universe Via Beirut 2 34014 Trieste Italy School of Physics Indian Institute of Science Education and Research Thiruvananthapuram Maruthamala PO Vithura Thiruvananthapuram 695551 Kerala India Physics Department Imperial College London Prince Consort Road SW7 2AZ London United Kingdom Centro Nazionale “High Performance Computer Big Data and Quantum Computing” Via Magnanelli 2 40033 Casalecchio di Reno (BO) Italy

We analyze the ringdown phase of the first detected black-hole merger, GW150914, using a simulation-based inference pipeline based on masked autoregressive flows. We obtain approximate marginal posterior distributions for the ringdown parameters, namely the mass, spin, and the amplitude and phases of the dominant mode and its first overtone. Thanks to the locally amortized nature of our method, we are able to calibrate our posteriors with injected simulations, producing posterior regions with guaranteed (i.e., exact) frequentist coverage of the true values. For GW150914, our calibrated posteriors provide only mild evidence (∼2σ) for the presence of an overtone, even if the ringdown is assumed to start at the peak of the amplitude.

关键词： Classical black holes General relativity Gravitational waves