检索结果-内蒙古大学图书馆

Interpretability Latent Space Method: Exploiting Shapley Representation to Explain Latent Space 7

学校读者我要写书评

暂无评论

Interpretability Latent Space Method: Exploiting Shapley Rep...

7th International Conference on Big Data and Information Analytics, BigDIA 2021

作者： Liu, Zitu Li, Jiawang Liu, Yue Liu, Qun Wang, Guoyin Guo, Yike School of Computer Engineering and Science Shanghai University Shanghai China Chongqing University of Posts and Telecommunications Chongqing Key Laboratory of Computational Intelligence Chongqing China Imperial College Department of Computing London United Kingdom

ISBN: (纸本)9781665424660

Shapley values have become one of the most popular interpretation methods for feature attribution. These methods attribute the prediction of the input by the machine learning model to its basic feature. However, most of the previous work is based on Shapley's interpretation after the training is completed because of Shapley's calculation requirements (exponential time complexity), so that they can focus on post-hoc Shapley explanations. Therefore, we suggest trying to use Shapley value itself as a latent representation in the deep model to guide model training. First, we extract the latent space of the generation model as input of Shapley network. Then, we utilize Shapley network, which provides layer-wise transformation in the same forward pass and produce Shapley representation. Further, we apply Shapley representation to the calculation of the new model to explain the training process of the model. Finally, the experimental results show on the two datasets of MNIST and Fashion MNIST that our proposed method has certain interpretability. © 2021 IEEE.

关键词： Game theory

Structure-Aware Generative Adversarial Network for Text-to-Image Generation

学校读者我要写书评

暂无评论

Structure-Aware Generative Adversarial Network for Text-to-I...

IEEE International Conference on Image Processing

作者： Wenjie Chen Zhangkai Ni Hanli Wang Department of Computer Science and Technology Tongji University Shanghai P. R. China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai P. R. China

Text-to-image generation aims at synthesizing photo-realistic images from textual descriptions. Existing methods typically align images with the corresponding texts in a joint semantic space. However, the presence of the modality gap in the joint semantic space leads to misalignment. Meanwhile, the limited receptive field of the convolutional neural network leads to structural distortions of generated images. In this work, a structure-aware generative adversarial network (SaGAN) is proposed for (1) semantically aligning multimodal features in the joint semantic space in a learnable manner; and (2) improving the structure and contour of generated images by the designed content-invariant negative samples. Experimental results show that SaGAN achieves over 30.1% and 8.2% improvements in terms of FID on the datasets of CUB and COCO when compared with the state-of-the-art approaches.

关键词：

Prescribed-Time Synchronization of Multiweighted and Directed Complex Networks

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Xu, Linlong Liu, Xiwei Department of Computer Science and Technology Tongji University The Key Laboratory of Embedded System and Service Computing Ministry of Education Shanghai201804 China

In this note, we study the prescribed-time (PT) synchronization of multiweighted and directed complex networks (MWDCNs) via pinning control. Unlike finite-time and fixed-time synchronization, the time for synchronization can be preset as needed, which is independent of initial values and parameters like coupling strength. First and foremost, we reveal the essence of PT stability by improper integral, L'Hospital rule and Taylor expansion theory. Many controllers established previously for PT stability can be included in our new model. Then, we apply this new result on MWDCNs as an application. The synchronization error at the prescribed time is discussed carefully, so, PT synchronization can be reached. The network topology can be directed and disconnected, which means that the outer coupling matrices (OCMs) can be asymmetric and not connected. The relationships between nodes are allowed to be cooperative or competitive, so elements in OCMs and inner coupling matrices (ICMs) can be positive or negative. We use the rearranging variables' order technique to combine ICMs and OCMs together to get the sum matrices, which can make a bridge between multiweighted and single-weighted networks. Finally, simulations are presented to illustrate the effectiveness of our theory. Copyright © 2022, The Authors. All rights reserved.

关键词： Synchronization

Auction Design for Value Maximizers with Budget and Return-on-spend Constraints

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lu, Pinyan Xu, Chenyang Zhang, Ruilong ITCS Shanghai University of Finance and Economics China Huawei TCS Lab China Shanghai Key Laboratory of Trustworthy Computing East China Normal University China Department of Computer Science and Engineering University at Buffalo United States

The paper designs revenue-maximizing auction mechanisms for agents who aim to maximize their total obtained values rather than the classical quasi-linear utilities. Several models have been proposed to capture the behaviors of such agents in the literature. In the paper, we consider the model where agents are subject to budget and return-on-spend constraints. The budget constraint of an agent limits the maximum payment she can afford, while the return-on-spend constraint means that the ratio of the total obtained value (return) to the total payment (spend) cannot be lower than the targeted bar set by the agent. The problem was first coined by [5]. In their work, only Bayesian mechanisms were considered. We initiate the study of the problem in the worst-case model and compare the revenue of our mechanisms to an offline optimal solution, the most ambitious benchmark. The paper distinguishes two main auction settings based on the accessibility of agents’ information: fully private and partially private. In the fully private setting, an agent’s valuation, budget, and target bar are all private. We show that if agents are unit-demand, constant approximation mechanisms can be obtained;while for additive agents, there exists a mechanism that achieves a constant approximation ratio under a large market assumption. The partially private setting is the setting considered in the previous work [5] where only the agents’ target bars are private. We show that in this setting, the approximation ratio of the single-item auction can be further improved, and a Ω(1/√n)approximation mechanism can be derived for additive agents. Copyright © 2023, The Authors. All rights reserved.

关键词： Additives

SecBCS: a secure and privacy-preserving blockchain-based crowdsourcing system

学校读者我要写书评

暂无评论

science China(Information sciences) 2020年第3期63卷 20-33页

作者： Chao LIN Debiao HE Sherali ZEADALLY Neeraj KUMAR Kim-Kwang Raymond CHOO Key Laboratory of Aerospace Information Security and Trusted Computing Ministry of EducationSchool of Cyber Science and Engineering Wuhan University State Key Laboratory of Cryptology College of Communication and Information University of Kentucky Department of Computer Science and Engineering Thapar University Department of Information Systems and Cyber Security and Department of Electrical and Computer Engineering University of Texas at San Antonio

A robust and scalable crowd management infrastructure is crucial in addressing operational challenges when deploying high-density sensors and actuators in a smart city. While crowdsourcing is widely used in crowd management, conventional solutions, such as Upwork and Amazon Mechanical Turk, generally depend on a trusted third-party platform. There exist several potential security concerns(e.g., sensitive leakage, single point of failure and unfair judgment) in such a centralized paradigm. Hence, a recent trend in crowdsourcing is to leverage blockchain(a decentralized ledger technology) to address some of the existing limitations. A small number of blockchain-based crowdsourcing systems(BCSs) with incentive mechanisms have been proposed in the literature, but they are generally not designed with security in mind. Thus, we study the security and privacy requirements of a secure BCS and propose a concrete solution(i.e., SecBCS)with a prototype implementation based on JUICE.

关键词： crowdsourcing blockchain secure privacy-preserving smart contract

Feedback Stability Under Mixed Gain and Phase Uncertainty

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Liang, Jiajin Zhao, Di Qiu, Li The Googol Technology Guangdong Shenzhen China The Department of Control Science and Engineering National Key Laboratory of Autonomous Intelligent Unmanned Systems Frontiers Science Center for Intelligent Autonomous Systems Tongji University Shanghai China The Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology Clear Water Bay Kowloon Hong Kong

In this study, we investigate the robust feedback stability problem for multiple-input-multiple-output linear time-invariant systems involving sectored-disk uncertainty, namely, dynamic uncertainty subject to simultaneous gain and phase constraints. This problem is thereby called a sectored-disk problem. Employing a frequency-wise analysis approach, we derive a fundamental static matrix problem that serves as a key component in addressing the feedback stability. The study of this matrix problem heavily relies on the Davis-Wielandt (DW) shells of matrices, providing a profound insight into matrices subjected to simultaneous gain and phase constraints. This understanding is pivotal for establishing a less conservative sufficient condition for the matrix sectored-disk problem, from which we formulate several robust feedback stability conditions against sectored-disk uncertainty. Finally, several conditions based on linear matrix inequalities are developed for efficient computation and verification of feedback robust stability against sectored-disk uncertainty. Copyright © 2024, The Authors. All rights reserved.

关键词： Linear time-invariant system

VHASR: A Multimodal Speech Recognition System With Vision Hotwords

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Hu, Jiliang Li, Zuchao Wang, Ping Ai, Haojun Zhang, Lefei Zhao, Hai Key Laboratory of Aerospace Information Security and Trusted Computing Ministry of Education School of Cyber Science and Engineering Wuhan University Wuhan China School of Computer Science Wuhan University Wuhan China School of Information Management Wuhan University Wuhan China Department of Computer Science and Engineering Shanghai Jiao Tong University China

The image-based multimodal automatic speech recognition (ASR) model enhances speech recognition performance by incorporating audio-related image. However, some works suggest that introducing image information to model does not help improving ASR performance. In this paper, we propose a novel approach effectively utilizing audio-related image information and set up VHASR, a multimodal speech recognition system that uses vision as hotwords to strengthen the model’s speech recognition capability. Our system utilizes a dual-stream architecture, which firstly transcribes the text on the two streams separately, and then combines the outputs. We evaluate the proposed model on four datasets: Flickr8k, ADE20k, COCO, and OpenImages. The experimental results show that VHASR can effectively utilize key information in images to enhance the model’s speech recognition ability. Its performance not only surpasses unimodal ASR, but also achieves SOTA among existing image-based multimodal ASR.1 Copyright © 2024, The Authors. All rights reserved.

关键词： Speech recognition

Vertical Federated Learning: Challenges, Methodologies and Experiments

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wei, Kang Li, Jun Ma, Chuan Ding, Ming Wei, Sha Wu, Fan Chen, Guihai Ranbaduge, Thilina School of Electrical and Optical Engineering Nanjing University of Science and Technology Nanjing210094 China Ministry of Education China Data61 CSIRO SydneyNSW2015 Australia Beijing100191 China The Shanghai Key Laboratory of Scalable Computing and Systems Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai200240 China The State Key Laboratory for Novel Software Technology Nanjing University Nanjing210023 China

Recently, federated learning (FL) has emerged as a promising distributed machine learning (ML) technology, owing to the advancing computational and sensing capacities of end-user devices, as well as the increasing concerns on users' privacy. As a special architecture in FL, vertical FL (VFL) is capable of constructing a hyper ML model by embracing sub-models from different types of clients. These sub-models are trained locally by vertically partitioned data with distinct attributes. Therefore, the design of VFL is fundamentally different from that of conventional FL, raising new and unique research issues. In this paper, we aim to discuss key challenges in VFL with effective solutions, and conduct experiments on real-life datasets to shed light on these issues. Specifically, we first propose a general framework on VFL, and highlight the key differences between VFL and conventional FL. Then, we discuss research challenges rooted in VFL systems from four aspects, i.e., security and privacy risks, expensive computation and communication costs, possible structural damage caused by model splitting, and system heterogeneity. Afterwards, we develop solutions to addressing the aforementioned challenges, and conduct extensive experiments to showcase the effectiveness of the proposed solutions. Copyright © 2022, The Authors. All rights reserved.

关键词： Federated learning

Non-Convex Potential Games for Finding Global Solutions to Sensor Network Localization

学校读者我要写书评

暂无评论

Non-Convex Potential Games for Finding Global Solutions to S...

European Control Conference (ECC)

作者： Gehui Xu Guanpu Chen Yiguang Hong Baris Fidan Thomas Parisini Karl H. Johansson Key Laboratory of Systems and Control Academy of Mathematics and Systems Science Beijing China Division of Decision and Control Systems School of Electrical Engineering and Computer Science KTH Royal Institute of Technology Stockholm Sweden Department of Control Science and Engineering Tongji University Shanghai China Shanghai Research Institute for Intelligent Autonomous Systems Shanghai China Department of Mechanical and Mechatronics Engineering University of Waterloo Waterloo ON Canada Department of Electrical and Electronic Engineering Imperial College London London UK Department of Engineering and Architecture University of Trieste Trieste Italy

ISBN: (数字)9783907144107

ISBN: (纸本)9798331540920

Sensor network localization (SNL) problems require determining the physical coordinates of all sensors in a network. This process relies on the global coordinates of anchors and the available measurements between non-anchor and anchor nodes. Attributed to the intrinsic non-convexity, obtaining a globally optimal solution to SNL is challenging, as well as implementing corresponding algorithms. In this paper, we formulate a non-convex multi-player potential game for a generic SNL problem to investigate the identification condition of the global Nash equilibrium (NE) therein, where the global NE represents the global solution of SNL. We employ canonical duality theory to transform the non-convex game into a complementary dual problem. Then we develop a conjugation-based algorithm to compute the stationary points of the complementary dual problem. On this basis, we show an identification condition of the global NE: the stationary point of the proposed algorithm satisfies a duality relation. Finally, simulation results are provided to validate the effectiveness of the theoretical results.

关键词： Location awareness Coordinate measuring machines Simulation Current measurement Noise Europe Games