检索结果-内蒙古大学图书馆

Robust Privacy-Preserving Federated Item Ranking in Online Marketplaces: Exploiting Platform Reputation for Effective Aggregation

引用

IEEE TRANSACTIONS ON BIG DATA 2025年第1期11卷 303-309页

作者： Ramos, Guilherme Boratto, Ludovico Marras, Mirko Univ Lisbon Dept Informat Inst Super Tecn P-1049001 Lisbon Portugal Inst Telecomunicacoes P-1049001 Lisbon Portugal Univ Cagliari Dept Math & Comp Sci I-09124 Cagliari Italy

Online marketplaces often collect products to sell from several other platforms (and sellers) and produce a unique ranking/score of these products to users. Keeping as private the user preferences provided in each (individual) platform is a need and a challenge at the same time. We are currently used to rating items in the marketplace itself which, in turn, can produce more effective rankings. Hence, the shaping of an effective item ranking would require a sharing of the user ratings between the individual platforms and the marketplace, thus impacting users' privacy. In this paper, we propose the initial steps towards a change of paradigm, where the ratings are kept as private in each platform. Under this paradigm, each platform produces its rankings, then aggregated by the marketplace, in a federated fashion. To ensure that the marketplace's rankings maintain their effectiveness, we exploit the concept of reputation of the individual platform, so that the final marketplace ranking is weighted by the reputation of each platform providing its ranking. Experiments on three datasets, covering different use cases, show that our approach can produce effective rankings, improving robustness to attacks, while keeping user preference data private within each seller platform.

关键词： Federated learning Data privacy Privacy Accuracy Training Data models computational modeling computational complexity Big Data Arithmetic privacy-preserving computing recommender systems reputation systems

来源：评论

学校读者我要写书评

暂无评论

Economic-Emission Coordinated Operation of Transportation-Microgrid Coupled System Based on Dynamic User Equilibrium

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2025年第2期26卷 2381-2396页

作者： Li, Bin Liu, Zhitao Su, Hongye Zhejiang Univ Inst Cyber Syst & Control State Key Lab Ind Control Technol Hangzhou 310027 Peoples R China Shenzhen Polytech Univ Inst Intelligence Sci & Engn Shenzhen 518055 Peoples R China

Booming in electric vehicles and the progress of dynamic wireless charging are deepening the interdependence of transportation networks (TNs) and microgrids (MGs). An imperative exists for a framework to coordinate dynamic traffic assignment (DTA) of the TN and optimal scheduling of MGs. Considering the simultaneous route-and-departure-time choice dynamic user equilibrium (DUE), this paper presents a noncooperative framework targeting the economic-emission coordinated operation of the transportation-microgrid coupled system. This framework is a Nash-Stackelberg-Nash game, where the leader problem is an economic-emission scheduling model of MGs and the follower problem is a DUE model of DTA for the TN. The best response algorithm is adopted to solve this framework. The DUE model is converted into an equivalent fixed-point problem, while the economic-emission scheduling model of each MG is reformulated as a multi-objective mixed-integer nonlinear program (MOMINLP). Moreover, we develop an alternating operator evolutionary algorithm (AOEA) to handle this MOMINLP. Case studies demonstrate that the proposed framework enables the coupled system to operate synergistically and AOEA has a competitive performance.

关键词： Costs Power system dynamics Vehicle dynamics Power systems Carbon dioxide Junctions computational modeling Vehicles Dynamic scheduling Optimal scheduling Transportation network dynamic traffic assignment dynamic user equilibrium microgrids electric vehicles

来源：评论

学校读者我要写书评

暂无评论

EM-Trans: Edge-Aware Multimodal Transformer for RGB-D Salient Object Detection

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2025年第2期36卷 3175-3188页

作者： Chen, Geng Wang, Qingyue Dong, Bo Ma, Ruitao Liu, Nian Fu, Huazhu Xia, Yong Northwestern Polytech Univ Sch Comp Sci & Engn Natl Engn Lab Integrated Aero Space Ground Ocean Xian 710072 Peoples R China Zhejiang Univ Ctr Brain Imaging Sci & Technol Hangzhou 310027 Peoples R China Mohamed Bin Zayed Univ Artificial Intelligence Comp Vis Dept Abu Dhabi U Arab Emirates ASTAR Inst High Performance Comp IHPC Singapore 138632 Singapore

RGB-D salient object detection (SOD) has gained tremendous attention in recent years. In particular, transformer has been employed and shown great potential. However, existing transformer models usually overlook the vital edge information, which is a major issue restricting the further improvement of SOD accuracy. To this end, we propose a novel edge-aware RGB-D SOD transformer, called EM-Trans, which explicitly models the edge information in a dual-band decomposition framework. Specifically, we employ two parallel decoder networks to learn the high-frequency edge and low-frequency body features from the low- and high-level features extracted from a two-steam multimodal backbone network, respectively. Next, we propose a cross-attention complementarity exploration module to enrich the edge/body features by exploiting the multimodal complementarity information. The refined features are then fed into our proposed color-hint guided fusion module for enhancing the depth feature and fusing the multimodal features. Finally, the resulting features are fused using our deeply supervised progressive fusion module, which progressively integrates edge and body features for predicting saliency maps. Our model explicitly considers the edge information for accurate RGB-D SOD, overcoming the limitations of existing methods and effectively improving the performance. Extensive experiments on benchmark datasets demonstrate that EM-Trans is an effective RGB-D SOD framework that outperforms the current state-of-the-art models, both quantitatively and qualitatively. A further extension to RGB-T SOD demonstrates the promising potential of our model in various kinds of multimodal SOD tasks.

关键词： Image edge detection Feature extraction Transformers Object detection computational modeling Task analysis Decoding Edge-aware model multimodal learning salient object detection (SOD) transformer

来源：评论

学校读者我要写书评

暂无评论

Activation-Guided Low-Rank Parameter Adaptation for Efficient Model Fine-Tuning

引用

IEEE ACCESS 2025年 13卷 70909-70918页

作者： Wang, Qingchen Shen, Shengyu Nanjing Yunwen Network Technol Co Ltd Zijin Res & Innovat Ctr Nanjing 210015 Jiangsu Peoples R China

Fine-tuning large language models is computationally expensive, and while existing parameter-efficient methods like Low-Rank Adaptation (LoRA) reduce computational costs, they are limited by suboptimal initialization strategies. We introduce Activation-Guided LoRA (AG-LoRA), a novel approach that initializes LoRA modules using Singular Value Decomposition (SVD) guided by activation patterns. Our method employs pre-trained weights combined with activation-based weighting factors and implements a new global rank assignment strategy that accounts for activation outliers. Experimental evaluations on LLaMA and CLIP models show that AG-LoRA achieves superior performance while reducing GPU memory usage compared to existing methods. In tests with LLaMA 7B models, AG-LoRA reached 75.9% accuracy across various tasks, surpassing both LoRA and DoRA baselines. AG-LoRA demonstrates significant improvements in parameter-efficient fine-tuning of large language models, offering enhanced performance and reduced computational requirements. These advances make it a promising solution for efficient model adaptation across diverse applications.

关键词： Adaptation models computational modeling Resource management Context modeling Training Accuracy Tuning Large language models Commonsense reasoning Vectors Deep learning low-rank adaptation parameter-efficient fine-tuning

来源：评论

学校读者我要写书评

暂无评论

MCECF: A Multiscale Complementary Enhanced Context Fusion Network for Remote Sensing Change Detection

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2025年 63卷

作者： Huang, Zhiyong Qiu, Hongjiang Hou, Mingyang Yu, Zhi Wang, Shiwei Li, Xiaoyu Wang, Jiahong Yan, Yan Liu, Yushi Chongqing Univ Sch Microelect & Commun Engn Chongqing 400044 Peoples R China Chongqing Univ Bioengn Coll Chongqing 400044 Peoples R China Chinese Peoples Liberat Army Gen Hosp Hlth Management Inst Med Ctr Natl Clin Res Ctr Geriatr Dis 2 Beijing 100853 Peoples R China Chinese Acad Sci Inst Adv Technol Shenzhen 518055 Peoples R China

Remote sensing change detection (RSCD) holds significant research value in remote sensing (RS) image processing. In recent years, many researchers have achieved remarkable results in RSCD tasks using methods based on convolutional neural networks (CNNs) or Transformers. Considering the limited receptive field of CNN models and the high computational cost of Transformers, many researchers have combined the two approaches, yielding promising results. However, most current RSCD-based models focus solely on change and temporal information, overlooking their complementary relationship. Additionally, some multiscale feature fusion methods emphasize enhancing individual scales while neglecting the correlations between different scales. To address the above issues, we propose a multiscale complementary enhanced context fusion (MCECF) network. The network first introduces a global-local context aggregation module (GLCAM) to capture global-local context information while extracting multilevel feature maps. Subsequently, a complementary enhancement difference module (CEDM) is employed to complementarily aggregate the captured change and temporal information of bi-temporal RS image features. To fully leverage the correlations between multiscale features, a progressive decoder comprising a supervised spatial attention (SSA) mechanism and a multiscale complementary enhanced fusion module (MCEFM) was developed. Moreover, to tackle the disparity between changed and unchanged regions, a dual-branch dynamic attention fusion module (DAFM) was designed to enhance the model's adaptability to diverse scenarios. We conducted comparative experiments on five RSCD datasets against nine state-of-the-art (SOTA) methods, and the results confirmed the effectiveness of the proposed MCECF in RSCD tasks. Our code will be made available at https://***/kakuqikaduo/MCECF

关键词： Remote sensing Feature extraction Transformers Decoding Complexity theory Data mining computational modeling Lighting Correlation Semantics Complementary enhancement feature fusion multiscale complementarity remote sensing change detection (RSCD)

来源：评论

学校读者我要写书评

暂无评论

Privacy-Preserving Recommendations With Mixture Model-Based Matrix Factorization Under Local Differential Privacy

引用

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS 2025年第7期21卷 5451-5459页

作者： Zhang, Pengfei Sun, Hong Zhang, Zhikun Cheng, Xiang Zhu, Youwen Zhang, Ji Anhui Univ Sci & Technol Sch Comp Sci & Engn State Key Lab Digital Intelligent Technol Unmanned Huainan 232001 Peoples R China Wuhan Text Univ Sch Econ Wuhan Peoples R China Zhejiang Univ Hangzhou 310058 Peoples R China Beijing Univ Posts & Telecommun Beijing 100876 Peoples R China Nanjing Univ Aeronaut & Astronaut Nanjing 210016 Peoples R China Univ Southern Queensland Toowoomba Qld 4350 Australia

Matrix factorization-based recommendations have emerged as a cornerstone for many industrial recommendation algorithms. However, studies employing differential privacy often rely on a trusted server, while those implementing local differential privacy (LDP) frequently encounter substantial accuracy degradation and excessive communication overhead due to noise injection and frequent user interactions. Moreover, the presence of injected noise for privacy protection and inherent Gaussian noise within these perturbed values compounds the accuracy issues and may create a cascading effect under LDP, exacerbating the complexity of the problem at hand. To address these challenges, we propose MENTOR, which is Mixture model-based rEcommeNdations approach with matrix FacTORization under LDP. Its main idea lies in the adoption of a bounded input perturbation mechanism that closely approximates the Laplace distribution while providing rigorous LDP, coupled with accounting for various noise types present in the perturbed data. This approach allows us to reformulate the matrix factorization process with minimal user interaction, requiring only a single round of communication. In particular, to add noise, we design a bidirectional bounded LDP input perturbation mechanism BBV while minimizing variance. To generate recommendation results, we devise a matrix factorization technique GLMF based on a Gaussian-Laplacian mixture model. Comprehensive experiments reveal that MENTOR outperforms the state of the art by at least 15% in RMSE and 12% in F-Measure, showcasing its effectiveness in balancing privacy and utility in recommendation systems.

关键词： Noise Privacy Perturbation methods Accuracy Recommender systems Gaussian noise Electronic mail Differential privacy Training computational modeling Local differential privacy (LDP) matrix factorization (MF) recommender system

来源：评论

学校读者我要写书评

暂无评论

TinyFL_HKD: Enhancing Edge AI Federated Learning With Hierarchical Knowledge Distillation Framework

引用

IEEE SENSORS JOURNAL 2025年第7期25卷 12038-12047页

作者： Hung, Chung-Wen Tsai, Cheng-Yu Wang, Chun-Chieh Lee, Ching-Hung Natl Yunlin Univ Sci & Technol Dept Elect Engn Yunlin 64002 Taiwan Natl Yang Ming Chiao Tung Univ Inst Elect & Control Engn Hsinchu 30010 Taiwan Chun Yuan Christian Univ Dept Elect Engn Taoyuan City 320314 Taiwan

With the rapid evolution of artificial intelligence (AI) and the Internet of Things (IoT), machine learning is increasingly being integrated into embedded systems, bringing computational capabilities closer to where data are generated. This article introduces a tiny federated learning framework, which concerns privacy, personalized training, and the constrained computational resources of edge platforms by introducing a novel hierarchical knowledge distillation (HKD), called TinyFL_HKD. The HKD leverages hierarchical learning and advanced encryption security (AES) schemes to ensure data privacy and security. It employs knowledge distillation to reduce model complexity for implementation in edge devices and enhance personalization. The performance of TinyFL_HKD is introduced by using two datasets: the tool wear dataset and the PHM 2010 Data Challenge dataset. Experimental results indicate that the HKD framework surpasses traditional federated averaging (FedAvg) and personalized federated learning (PFL) algorithms in both model accuracy and computational efficiency. This establishes HKD as a resilient solution for edge AI applications.

关键词： computational modeling Training Artificial intelligence Data privacy Vectors Random access memory Nonvolatile memory Ions Ferroelectric films Servers encryption federated learning hierarchical learning

来源：评论

学校读者我要写书评

暂无评论

A Weighted Deep Learning-Based Predictive Control for Multimode Nonlinear System With Industrial Applications

引用

IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING 2025年 22卷 10814-10826页

作者： Huang, Keke Cao, Wenpu Liu, Yishun Wu, Dehao Yang, Chunhua Gui, Weihua Cent South Univ Sch Automat Changsha 410083 Peoples R China

In response to the challenge of strongly nonlinear and multimode systems control, this paper introduces a weighted deep learning based adaptive predictive control method. This approach integrates LSTM networks for different operating modes using a set of weighting coefficients. These coefficients are dynamically updated during online control via an error-guided scheduling strategy to adapt to changing operation modes. Compared to offline identification based methods, the proposed method eliminates the need for mode recognition or model switching strategies and can adapt to drifted operation modes. In contrast to online methods, it achieves rapid model convergence and reduced computational cost, requiring only minimal data to update the weighting coefficients without necessitating the retraining of the LSTM networks. Theoretical convergence and stability analysis ensure the reliability of the proposed method. Numerical simulations and industrial control experiments demonstrate that the proposed approach exhibits favorable control performance across both known and drifted operation modes. Note to Practitioners-Considering the changing operation modes in complex industrial processes and the detrimental effect of slow or unstable control during system operation, this paper proposes a weighted LSTM based predictive control method for strongly nonlinear and multimode systems. Extensive experiments demonstrate that compared to other state-of-the-art methods, this method can rapidly adapt to changes in operating modes with a small amount of data, meeting both real-time and stability requirements for online control.

关键词： Predictive models Long short term memory Adaptation models Mathematical models Convergence computational modeling Predictive control Optimization Computer architecture Switches Multimode control weighted networks deep learning long short-term memory network model predictive control

来源：评论

学校读者我要写书评

暂无评论

Digital Twin-Aided Vehicular Edge Network: A Large-Scale Model Optimization by Quantum-DRL

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2025年第2期74卷 2156-2173页

作者： Paul, Anal Singh, Keshav Li, Chih-Peng Dobre, Octavia A. Duong, Trung Q. Natl Sun Yat Sen Univ Inst Commun Engn Kaohsiung 80424 Taiwan Mem Univ Fac Engn & Appl Sci St John NF A1C 5S7 Canada Queens Univ Belfast Sch Elect Elect Engn & Comp Sci Belfast BT7 1NN North Ireland Kyung Hee Univ Dept Elect Engn Yongin 17104 South Korea

This paper presents an innovative large model framework for optimizing the task offloading efficiency in vehicular edge networks, with a focus on ultra-reliable low-latency communication. We introduce a comprehensive model that integrates quantum computing with a deep reinforcement learning (DRL) model, supported by long short-term memory (LSTM) networks and a digital twin framework. This integration is designed to address the complexities of distributed vehicular edge computing networks, targeting efficient latency, energy, and quality-of-service management. Our model utilizes the parallel processing capabilities of quantum computing to enhance the DRL algorithm, effectively handling high-dimensional decision spaces. LSTM networks provide predictive insights into future network states in a digital twin framework and ensure real-time synchronization and adaptive strategy optimization. We employ a multi-agent framework, encompassing vehicles, unmanned aerial vehicles, and base stations, each utilizing a Nash equilibrium-based strategy for optimal decision-making, supplemented by incentive and penalty functions for reward optimization. Simulation results demonstrate notable improvements in task offloading efficiency, highlighting the model's efficacy over conventional DRL models.

关键词： Task analysis computational modeling Quantum computing Optimization Ultra reliable low latency communication Edge computing Data models Vehicular edge computing task offloading ultra-reliable low-latency communication quantum deep reinforcement learning large model framework long short-term memory networks digital twin and multi-agent systems

来源：评论

学校读者我要写书评

暂无评论

FreqWave-TranDuD: A Multivariate Time Series Anomaly Detection Method Based on Wavelet and Fourier Transforms

引用

IEEE ACCESS 2025年 13卷 68384-68397页

作者： Huang, Yashi Liu, Peishun Han, Rongjia Yin, Tian Dou, Quanjie Song, Yibao Luo, Mengqi Ocean Univ China Fac Informat Sci & Engn Qingdao 266100 Shandong Peoples R China

Time series anomaly detection (AD) plays a crucial role in network systems. It enables the timely detection of anomalies and root cause analysis, helping to prevent unnecessary losses. Existing methods have not fully exploited the frequency information embedded in time series data, limiting their ability to effectively capture global and periodic patterns. In this paper, we propose a novel time series anomaly detection method based on frequency domain feature extraction, referred to as FreqWave-TranDuD. The method adopts an encoder-decoder deep learning architecture, integrating Fourier Transform and Wavelet Transform to extract comprehensive time-frequency information. Fast Fourier Transform (FFT) is used to capture global periodic patterns, while Wavelet Transform captures local patterns. Additionally, LSTM is employed to capture temporal dependencies in the time series. The extracted features are then preliminarily concatenated and fed into the encoder-decoder architecture for accurate anomaly detection. Extensive experiments on six public datasets from domains such as aerospace detection and water treatment demonstrate that our method, FreqWave-TranDuD, outperforms other advanced baseline models.

关键词： Time series analysis Anomaly detection Feature extraction Wavelet transforms Time-frequency analysis Transformers computational modeling Fast Fourier transforms Long short term memory Generative adversarial networks Deep learning unsupervised learning anomaly detection FFT Wavelet Transform

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：