检索结果-内蒙古大学图书馆

Machine Unlearning: Taxonomy, Metrics, Applications, Challenges, and Prospects

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Na Zhou, Chunyi Gao, Yansong Chen, Hui Fu, Anmin Zhang, Zhi Yu, Shui School of Cyberspace Science and Engineering Nanjing University of Science and Technology Nanjing210094 China State Key Laboratory of Integrated Services Networks Xidian University Xi’an710071 China School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210094 China Data61 CSIRO CanberraACT2601 Australia Department of Computer Science and Software Engineering University of Western Australia PerthWA6009 Australia School of Computer Science University of Technology Sydney SydneyNSW2007 Australia

Personal digital data is a critical asset, and governments worldwide have enforced laws and regulations to protect data privacy. data users have been endowed with the 'right to be forgotten' of their data. In the course of machine learning (ML), the forgotten right requires a model provider to delete user data and its subsequent impact on ML models upon user requests. Machine unlearning emerges to address this, which has garnered ever-increasing attention from both industry and academia. While the area has developed rapidly, there is a lack of comprehensive surveys to capture the latest advancements. Recognizing this shortage, we conduct an extensive exploration to map the landscape of machine unlearning including the (fine-grained) taxonomy of unlearning algorithms under centralized and distributed settings, debate on approximate unlearning, verification and evaluation metrics, challenges and solutions for unlearning under different applications, as well as attacks targeting machine unlearning. The survey concludes by outlining potential directions for future research, hoping to serve as a guide for interested scholars. Copyright © 2024, The Authors. All rights reserved.

关键词： Taxonomies

Personalized Federated Learning with Collaborative Aggregation Networks for Multi-Site Brain Disorder Diagnosis

学校读者我要写书评

暂无评论

Personalized Federated Learning with Collaborative Aggregati...

Industrial Automation, Robotics and Control engineering (IARCE), International Conference on

作者： Qian Si Yang Li School of Cyber Science and Technology Beihang University Beijing China Department of Automation Science and Electrical Engineering Beijing Advanced Innovation Center for Big Data and Brain Computing State Key Laboratory of Virtual Reality Technology and Systems Advanced Institute of Information Technology Peking University Beihang University Beijing China

ISBN: (数字)9798350380323

ISBN: (纸本)9798350380330

In multi-site brain disease diagnosis studies, traditional centralized training methods necessitate sharing medical data, posing significant privacy risks. Federated learning (FL) offers a privacy-preserving solution by enabling global model training through aggregating locally trained models from multiple data centers without sharing raw data. However, current FL approaches rely on a server-based network topology, where central server failure disrupts training. Additionally, data heterogeneity across sites often slows convergence and reduces accuracy. To overcome these issues, we introduce a decentralized personalized federated learning collaborative aggregation network (pFedCAN). This framework has two core components: (1) separating local models into shared and personalized layers, and (2) forming a collaborative aggregation network via similarity detection in the shared layers. Specifically, each center trains its local model, then separates it into shared and personalized layers. The shared layer is exchanged with other centers, while the personalized layer remains local. data centers analyze similarities in received shared layers to build a collaborative network, where shared layers from similar centers are aggregated to refine the model. This approach flexibly adapts to varying levels of data heterogeneity, enhancing model training efficiency. Validation on public datasets, ABIDE I and ADHD, shows that the proposed method outperforms current leading techniques.

关键词： Training Adaptation models data centers Federated learning Service robots Collaboration Brain modeling data models Servers Medical diagnostic imaging

Rapid Estimation and Pattern Discovery of Dynamic Vehicle Emissions on Road Networks Based on the Traffic Performance Index

学校读者我要写书评

暂无评论

SSRN

SSRN 2025年

作者： Zang, Jinrui Hu, Xin Shi, Kedi Song, Guohua Li, Zhihong Beijing Advanced Innovation Center for Future Urban Design Beijing University of Civil Engineering and Architecture 1 Zhanlanguan Road Xicheng District Beijing100044 China Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport Beijing Jiaotong University 3 Shangyuancun Haidian District Beijing100044 China

The estimation of dynamic emissions from mobile sources such as vehicles presents a major challenge in air quality modeling. Current studies indicate a strong positive correlation between traffic congestion and emission levels. However, there is a notable lack of research defining the relationship quantitatively. The dynamic and rapid estimation of total emissions from road networks holds critical importance using real-time Traffic Performance Indices (TPIs). This paper focuses on rapidly estimating and identifying patterns in dynamic vehicle emissions across road networks by leveraging TPIs. At first, 1the relationship between vehicle emissions and TPIs is rigorously examined. Subsequently, a quantitative sine function model is developed to characterize the relationship. Finally, employing the SOM neural network algorithm, temporal emission patterns are investigated, and the impact of the COVID-19 pandemic on emissions is assessed. The results demonstrate: 1) A non-linear positive correlation exists between total emissions and TPIs with emissions escalating with increasing TPI;2) The proposed sine function model accurately predicts total network emissions with average relative errors of 9.26% for NOx, 9.04% for HC, 8.86% for CO2, and 9.02% for CO;3) Utilizing the Self-Organizing Map (SOM) neural network clustering algorithm identifies eight emission variation patterns including Festivals, Saturdays, Sundays, Workdays during vacations, Mondays, Fridays, Ordinary workdays, and Congested workdays, which effectively represent 91.54% of emission scenarios across the network. The findings are expected to broaden TPI applications while providing robust scientific support for integrated policies targeting congestion management and pollution control. © 2025, The Authors. All rights reserved.

关键词： Traffic congestion

Adaptive feature normalization network for single image dehazing

学校读者我要写书评

暂无评论

Adaptive feature normalization network for single image deha...

2021 International Conference on Advanced Algorithms and Signal Image Processing, AASIP 2021

作者： Hu, Guangrui Yang, Yueting Yang, Chuansheng Wang, Chao Tan, Anhui School of Information Engineering Zhejiang Ocean University Zhoushan316022 China School of Mathematics and Statistics Beihua University Jilin Jilin132013 China Key Laboratory of Oceanographic Big Data Mining and Application of Zhejiang Province Zhoushan316022 China

In this paper, we design an end-to-end adaptive feature normalization network (AFN-Net) for single image dehazing. In order to fit the function that can recover haze-free images from haze images more effectively, an Adaptive Normalization module (AN) is designed, which can unify the input into the same feature space. A large number of experimental evaluations show that the proposed method is superior to the state-of-the-art (SOTA) method on the benchmark datasets. © 2021 Institute of Physics Publishing. All rights reserved.

关键词： Large dataset

Multi-Objective Genetic Programming Based on Decomposition for Feature Learning in Image Classification

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Zhang, Tuo Bi, Ying Liang, Jing Xue, Bing Zhang, Mengjie School of Electrical and Information Engineering Zhengzhou University 100 Kexue Rd Henan Zhengzhou450001 China State Key Laboratory of Intelligent Agricultural Power Equipment 39 Xiyuan Rd Henan Luoyang471039 China School of Electrical Engineering and Automation Henan Institute of Technology 699 Pingyuan Rd Henan Xinxiang453003 China School of Engineering and Computer Science Centre for Data Science and Artificial Intelligence Victoria University of Wellington PO Box 600 Wellington6140 New Zealand

Image classification presents a formidable challenge due to its high dimensionality and extensive variations. Feature learning is a powerful method in addressing this challenge, constituting a multi-objective problem aimed at maximizing classification accuracy and minimizing the number of learned features. A few multi-objective genetic programming (MOGP) methods have been proposed to optimize these two objectives, simultaneously. However, existing MOGP methods ignore the characteristics of feature learning tasks. Therefore, this work proposes a decomposition-based MOGP approach with a global replacement strategy for feature learning in data-efficient image classification. To handle the different value ranges of the two objectives, a transformation function is designed to uniform the range of the number of learned features. In addition, a preference-based decomposition strategy is proposed to address the preference for the objective of classification accuracy. The proposed approach is compared with existing MOGP methods for feature learning on five different image classification datasets with different numbers of training images. The experimental results demonstrate the effectiveness of the proposed approach by achieving a performance better than or comparable to the compared methods. © 2024, The Authors. All rights reserved.

关键词： Contrastive Learning

2D-DOA for a Monostatic ULA EMVS-MIMO Radar Based on RC-PARAFAC

学校读者我要写书评

暂无评论

IEEE Sensors Journal 2025年第12期25卷 21763-21774页

作者： Wang, Jianlong Shi, Junpeng Wen, Fangqing China Three Gorges University Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hvdroelectric Engineering College of Computer and Information Technology Yichang443002 China National University of Defense Technology College of Electronic Countermeasure Hefei230037 China City University of Macau College of Computer and Information Technology China Three Gorges University Faculty of Data Science 999078 China

The electromagnetic vector sensor (EMVS) embedded multiple-input multiple-output (MIMO) radar is an emerging technology capable of two-dimensional (2D) direction of arrival (DOA) estimation. In this paper, we propose a low-complexity algorithm for monostatic uniform linear array (ULA) EMVS-MIMO radar to rapidly estimate 2D-DOA. First, we decompose the received signal into a third-order tensor using parallel factor analysis (PARAFAC). Next, we eliminate data redundancy through a reduced complexity (RC) process. Then, we use the least squares method to extract ambiguous elevation angle information from the spatial response matrix and apply the vector cross-product (VCP) technique to the electromagnetic vector to achieve unambiguous 2D-DOA estimation. Finally, we combine these estimates to achieve high-resolution, unambiguous elevation angle estimation. The proposed algorithm is suitable for large-scale and sparse EMVS-MIMO radar systems. © 2001-2012 IEEE.

关键词： Direction of arrival

Style Content Decomposition-based data Augmentation for Domain Generalizable Medical Image Segmentation

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Shen, Zhiqiang Cao, Peng Yang, Jinzhu Zaiane, Osmar R. Chen, Zhaolin School of Computer Science and Engineering Northeastern University Shenyang China Key Laboratory of Intelligent Computing in Medical Image Ministry of Education Shenyang China Alberta Machine Intelligence Institute University of Alberta EdmontonAB Canada Monash Biomedical Imaging Monash University Australia Department of Data Science and AI Monash University Australia

Due to the domain shifts between training and testing medical images, learned segmentation models often experience significant performance degradation during deployment. In this paper, we first decompose an image into its style code and content map and reveal that domain shifts in medical images involve: style shifts (i.e., differences in image appearance) and content shifts (i.e., variations in anatomical structures), the latter of which has been largely overlooked. To this end, we propose StyCona, a style content decomposition-based data augmentation method that innovatively augments both image style and content within the rank-one space, for domain generalizable medical image segmentation. StyCona is a simple yet effective plug-and-play module that substantially improves model generalization without requiring additional training parameters or modifications to the segmentation model architecture. Experiments on cross-sequence, cross-center, and cross-modality medical image segmentation settings with increasingly severe domain shifts, demonstrate the effectiveness of StyCona and its superiority over state-of-the-arts. The code is available at https://***/Senyh/StyCona. Copyright © 2025, The Authors. All rights reserved.

关键词： Medical imaging

Optimal Bilinear Equalizer for Cell-Free Massive MIMO Systems over Correlated Rician Channels

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Zhe Zhang, Jiayi Björnson, Emil Niyato, Dusit Ai, Bo State Key Laboratory of Advanced Rail Autonomous Operation China School of Electronics and Information Engineering Beijing Jiaotong University Beijing100044 China Department of Computer Science KTH Royal Institute of Technology Stockholm114 28 Sweden College of Computing & Data Science Nanyang Technological University 639798 Singapore

In this paper, we explore the low-complexity optimal bilinear equalizer (OBE) combining scheme design for cell-free massive multiple-input multiple-output networks with spatially correlated Rician fading channels. We provide a spectral efficiency (SE) performance analysis framework for both the centralized and distributed processing schemes with bilinear equalizer (BE)-structure combining schemes applied. The BE-structured combining is a set of schemes that are constructed by the multiplications of channel statistics-based BE matrices and instantaneous channel estimates. Notably, we derive closed-form achievable SE expressions for centralized and distributed BE-structured combining schemes. We propose one centralized and two distributed OBE schemes: Centralized OBE (C-OBE), Distributed OBE based on Global channel statistics (DG-OBE), and Distributed OBE based on Local channel statistics (DL-OBE), which maximize their respective SE expressions. OBE matrices in these schemes are tailored based on varying levels of channel statistics. Notably, we obtain new and insightful closed-form results for the C-OBE, DG-OBE, and DL-OBE combining schemes. Numerical results demonstrate that the proposed OBE schemes can achieve excellent SE, even in scenarios with severe pilot contamination. Copyright © 2024, The Authors. All rights reserved.

关键词： Fading channels

6D Movable Antenna Enhanced Interference Mitigation for Cellular-Connected UAV Communications

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Ren, Tianshi Zhang, Xianchao Zhu, Lipeng Ma, Wenyan Gao, Xiaozheng Zhang, Rui School of Information and Electronics Beijing Institute of Technology Beijing100081 China Department of Electrical and Computer Engineering National University of Singapore Singapore117583 Singapore Provincial Key Laboratory of Multimodal Perceiving and Intelligent Systems Jiaxing University Jiaxing314001 China Engineering Research Center of Intelligent Human Health Situation Awareness of Zhejiang Province Jiaxing University Jiaxing314001 China School of Science and Engineering Shenzhen Research Institute of Big Data The Chinese University of Hong Kong Guangdong Shenzhen518172 China

Cellular-connected unmanned aerial vehicle (UAV) communications is an enabling technology to transmit control signaling or payload data for UAVs through cellular networks. Due to the line-of-sight (LoS) dominant air-to-ground channels, efficient interference mitigation is crucial to UAV communications, while the conventional fixed-position antenna (FPA) arrays have limited degrees of freedom (DoFs) to suppress the interference between the UAV and its non-associated co-channel base stations (BSs). To address this challenge, we propose in this letter a new approach by utilizing the six-dimensional movable antenna (6DMA) arrays to enhance the interference mitigation for the UAV. Specifically, we propose an efficient block coordinate descent (BCD) algorithm to iteratively optimize the antenna position vector (APV), array rotation vector (ARV), receive beamforming vector, and associated BS of the UAV to maximize its signal-to-interference-plus-noise ratio (SINR). Numerical results show that the proposed 6DMA enhanced cellular-connected UAV communication can significantly outperform that with the traditional FPA arrays and other benchmark schemes in terms of interference mitigation. Copyright © 2024, The Authors. All rights reserved.

关键词： Signal to noise ratio