检索结果-内蒙古大学图书馆

IEEE Transactions on Radar Systems 2025年 3卷 220-232页

作者： Wang, Zeyu Chen, Hongmeng Xu, Shuwen Li, Ming The State Key Laboratory of Networking and Switching Technology School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing100876 China Beijing Institute of Radio Measurement Beijing100854 China The National Laboratory of Radar Signal Processing Xidian University Shaanxi Xi’an710071 China

The weight vector in space–time adaptive processing (STAP) algorithm will lead to notches at the position of the interfering targets when there are interfering targets in the training data. If these interfering targets are close to the target of interest on the space–time spectrum, the target signal self-nulling occurs. To deal with this problem, a machine learning-aided nonhomogeneity detection (ML-NHD) method is proposed. More specifically, the subaperture smoothing technique is first performed on each training data to obtain the subaperture sample covariance matrices (SCMs). We prove that when the airborne radar works in side-looking mode and the clutter foldover factor is an integer, the numbers of large eigenvalues (EIGs) of the subaperture SCMs are different for the ordinary training data samples and outlier training data samples. Then, four features are constructed based on the differences in the characteristics of EIGs and eigenvectors of the subaperture SCMs. Finally, a binary classifier based on support vector machine (SVM) is trained to classify the ordinary training data and the outlier training data. The performance assessment shows that the ML-NHD method can detect the outlier training data effectively and achieves better performance of clutter suppression compared with the conventional methods. © 2025 IEEE. All rights reserved.

关键词： Eigenvalues and eigenfunctions

来源：评论

学校读者我要写书评

暂无评论

Soft Knowledge Distillation with Multi-Dimensional Cross-Net Attention for Image Restoration Models Compression

Soft Knowledge Distillation with Multi-Dimensional Cross-Net...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yongheng Zhang Danfeng Yan State Key Laboratory of Networking and Switching Technology BUPT Beijing China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Transformer-based encoder-decoder models have achieved remarkable success in image-to-image transfer tasks, particularly in image restoration. However, their high computational complexity—manifested in elevated FLOPs and parameter counts—limits their application in real-world scenarios. Existing knowledge distillation methods in image restoration typically employ lightweight student models that directly mimic the intermediate features and reconstruction results of the teacher, overlooking the implicit attention relationships between them. To address this, we propose a Soft Knowledge Distillation (SKD) strategy that incorporates a Multi-dimensional Cross-net Attention (MCA) mechanism for compressing image restoration models. This mechanism facilitates interaction between the student and teacher across both channel and spatial dimensions, enabling the student to implicitly learn the attention matrices. Additionally, we employ a Gaussian kernel function to measure the distance between student and teacher features in kernel space, ensuring stable and efficient feature learning. To further enhance the quality of reconstructed images, we replace the commonly used L1 or KL divergence loss with a contrastive learning loss at the image level. Experiments on three tasks—image deraining, deblurring, and denoising—demonstrate that our SKD strategy significantly reduces computational complexity while maintaining strong image restoration capabilities.

关键词： Representation learning Image coding Computational modeling Contrastive learning Signal processing Transformers Kernel Computational complexity Speech processing Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Dynamic Staleness Control for Asynchronous Federated Learning in Decentralized Topology 18th

Dynamic Staleness Control for Asynchronous Federated Learni...

引用

18th International Conference on Wireless Artificial Intelligent Computing Systems and Applications, WASA 2024

作者： Ma, Qianpiao Liu, Jianchun Jia, Qingmin Zhou, Xiaomao Hu, Yujiao Xie, Renchao Purple Mountain Laboratories Jiangsu Nanjing China School of Computer Science and Technology University of Science and Technology of China Anhui Hefei China State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9783031714665

Decentralized federated learning (DFL) has emerged as a promising paradigm for distributed machine learning over edge nodes (i.e., workers) without relying on a centralized parameter server. Most existing DFL researches rely on synchronous communication among workers. However, due to edge heterogeneity and dynamic network conditions, synchronous DFL mechanisms may suffer from inefficient model training and poor scalability. Meanwhile, the existing asynchronous DFL (ADFL) mechanisms present the challenge of stale models among workers, leading to diminished training quality, especially on Non-IID data. In this paper, we propose a novel staleness-aware ADFL (SA-ADFL) mechanism, aiming to realize a trade-off between model training efficiency and quality by dynamic staleness control. Specifically, we provide rigorous theoretical proof for SA-ADFL and formulate the worker scheduling problem to minimize total model training time under flexible long-term staleness constraints. Then we decompose the original round-coupled problem into a series of single-round sub-problems by leveraging the Lyapunov optimization, enabling efficient worker selection to minimize training time in each round while ensuring staleness queue stability. Experimental results demonstrate that our SA-ADFL accelerates model training by approximately 52.9% while maintaining equivalent model training accuracy compared with the state-of-the-art mechanisms. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

LEP: Leveraging Local Entropy Pruning for Sparsity in Large Language Models

LEP: Leveraging Local Entropy Pruning for Sparsity in Large ...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yuli Chen Bo Cheng Yingying Zhang Shuhao Zhang Zhixuan Wu Fanshen Meng State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

The application of Large Language Models (LLMs) is rapidly expanding in fields such as natural language processing and computer vision. However, due to the enormous number of model parameters, while their emergent capabilities enhance performance, they also incur significant computational and storage costs, particularly during inference. This makes it particularly challenging to deploy large models in resource-constrained environments. To address this issue, we propose an innovative pruning method based on local entropy, aimed at improving model performance while reducing computational costs. Our approach leverages a small amount of calibration data and calculates the importance of weights based on the local entropy of weights and input activations, applying it to both unstructured and semi-structured pruning without the need for complex weight updates or retraining. Experimental results demonstrate that our method shows superior performance across models of various scales. Specifically, at a sparsity rate of 50%, our method significantly reduces model perplexity without the need for weight updates and outperforms existing pruning techniques such as SparseGPT and Wanda in zero-shot tasks. The code is available at: this https URL.

关键词： Uniform resource locators Measurement Costs Computational modeling Large language models Signal processing Entropy Natural language processing Speech processing Hardware acceleration

来源：评论

学校读者我要写书评

暂无评论

Dynamic Delay-Doppler-Angle Domain Channel Tracking for THz massive MIMO-OTFS Communication Systems

引用

IEEE Transactions on Wireless Communications 2025年

作者： Fan, Yuheng Yang, Chuang Yuan, Renzhi Peng, Mugen Beijing University of Posts and Telecommunications State Key Laboratory of Networking and Switching Technology Beijing100876 China

As data transmission demand grows rapidly, terahertz (THz) frequency has been regarded as the most promising candidate in future sixth-generation (6G) communication systems. Facing severe path loss, THz communications have to use massive MIMO technology to promote the directional transmission gain in narrow beams. However, narrow beams and wide bandwidth signal in THz communications together result in the time-frequency-space non-stationary channel. To track this non-stationary channel, we first combine THz massive MIMO communication system with orthogonal time frequency space (OTFS) modulation to get the union sparse channel on delay-Doppler-angle (DDA) domain. Then this non-stationary channel is modeled through our proposed DDA-2D-Markov models (DDA-2D-MM) considering the unique channel characteristics on delay, Doppler and angle dimension respectively. Besides, we design a unique DDA-structured approximate message passing (DDA-STAMP) algorithm to track the dynamic channel with the prior information in consecutive time slots to improve the algorithm convergence speed and accuracy performance. Simulation results show that the proposed DDA-STAMP algorithm enhances both of the time consumption and accuracy performance in THz massive MIMO-OTFS communication channel tracking. © 2002-2012 IEEE.

关键词： Markov processes

来源：评论

学校读者我要写书评

暂无评论

Fed-MoE: Efficient Federated Learning for Mixture-of-Experts Models via Empirical Pruning 25th

Fed-MoE: Efficient Federated Learning for Mixture-of-Expert...

引用

25th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2024

作者： Zou, Yifei Qi, Senmao Yuan, Yuan Wang, Dawei Shen, Shikun Wu, Libing Guo, Shaoyong Yu, Dongxiao School of Computer Science and Technology Shandong University Qingdao China School of Cyber Science and Engineering Wuhan University Wuhan China State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9789819642069

In recent years, the combination of Federated Learning (FL) framework and Mixture of Experts (MoE) architecture has shown promise for large-scale model pre-training while preserving data privacy. However, the significant computational and communication requirement of MoE-based models presents substantial challenges for FL systems. To address this, we propose Fed-MoE, an efficient federated learning framework designed for MoE models through empirical pruning. Fed-MoE reduces computational overhead by pruning multiple experts at the client side while retaining general components, thus converting the model into dense sub-models. These sub-models are locally trained and later aggregated by the parameter server into a sparse MoE model. Our method optimizes the use of sparse activation in MoE models and reduces resource consumption without compromising performance. Extensive experimental results demonstrate that Fed-MoE significantly improves the efficiency of FL without compromising model performance, offering an effective solution for deploying large-scale MoE models in federated learning environments. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Predictive Beamforming for Joint Radar and Communication Transmission in LEO Satellite Aeronautical Systems

引用

IEEE Transactions on Vehicular technology 2025年

作者： Xv, Hongtao Sun, Yaohua Peng, Mugen Beijing University of Posts and Telecommunications State Key Laboratory of Networking and Switching Technology Beijing100876 China

LEO satellite communication is essential to provide aviation terminal (AT) services. However, due to the relative motion between LEO satellites and ATs, ATs are required to frequently report their current positions to the satellite to guarantee continuous service, which may lead to significant signaling overhead. To overcome this issue, we propose a predictive beamforming approach based on location sensing, whose core idea is to leverage a unified beampattern to enable sensing and data transmission simultaneously. Specifically, Cramér-Rao lower bound (CRLB) is adopted to characterize position tracking accuracy under the extended Kalman filtering framework. Then, the concerned problem is formulated to optimize spectral efficiency under transmission power and tracking accuracy constraints. To deal with the problem, we first prove the monotonic relationship between tracking accuracy and the perceived signal-to-interference-plus-noise ratio at the satellite to reformulate the problem, and then it is solved optimally by combining branch-and-reduce-and-bound approach with second-order cone programming. Towards practical application, the primal problem is equivalently reformulated with quadratic transform, and then semidefinite relaxation is adopted for low-complexity beamformer design. Simulation results verify the superior tracking and communication performance of our low-complexity design, and it is shown that it achieves comparable transmission rate and tracking accuracy relative to optimum. Moreover, the average spectral efficiency and tracking accuracy can be improved by up to 60% and 30% relative to other baselines. © 1967-2012 IEEE.

关键词： Extended Kalman filters

来源：评论

学校读者我要写书评

暂无评论

Beam Footprint Design, Scheduling, and Spectrum Assignment in Low Earth Orbit Mega-Constellations

引用

IEEE Transactions on Vehicular technology 2025年

作者： Zhu, Jianfeng Sun, Yaohua Feng, Xin'ao Wang, Shangguang Liu, Liang Peng, Mugen Beijing University of Posts and Telecommunications State Key Laboratory of Networking and Switching Technology Beijing100876 China

Owing to the advantages of ubiquitous connectivity and high capacity, low earth orbit mega-constellations have drawn much attention. However, it is challenging to maximize network throughput under much available beam resource, complex inter-beam interference, and spatially uneven traffic demand with low computation complexity. In this paper, we formulate a novel beam management problem including beam footprint design, beam scheduling, and spectrum assignment. Considering the tight coupling among them, we first develop a fast beam footprint design method for earth-fixed cells, which adapts to dynamic satellite attitudes. Then, with the reformulation of objective and constraints, the primal problem is transformed into a beam scheduling and spectrum assignment problem with much smaller solution space, and an optimum algorithm is designed by combining branch-and-bound and interior point methods. Towards practical application, a more scalable graph-based heuristic algorithm is proposed. Simulation results show that the developed beam footprint design method can significantly reduce inter-beam interference and improve network throughput, and the proposed heuristic algorithm achieves comparable throughput but reduces execution time by two orders of magnitude compared with the optimum one. Moreover, in scenarios with 120 cells, our proposal increases network throughput by over 60% compared to the baseline based on integer recovery. © 2025 IEEE.

关键词： Heuristic algorithms

来源：评论

学校读者我要写书评

暂无评论

Coordinated Multi-Point Enabled ISAC Under Asynchronous Errors: Performance Analysis and Waveform-Beamforming Optimization

引用

IEEE Transactions on Vehicular technology 2025年

作者： Zhang, Jiupeng Yan, Shi Peng, Mugen Ouyang, Qiu Beijing University of Posts and Telecommunications State Key Laboratory of Networking and Switching Technology Beijing100876 China

The coordinated multi-point approach presents promising opportunities for integrated sensing and communication, offering enhanced coverage and accuracy. However, inevitable timing inconsistencies across participating transceivers lead to signal asynchrony, diminishing the effectiveness of this paradigm. This paper presents a novel coordinated sensing and communication scheme designed to mitigate asynchronous challenges. Specifically, asynchronous errors are jointly estimated alongside target locations, then enabling synchronization corrections for improved coordinated communication. To quantify performance bounds, closed-form expressions of the Cramer-Rao bound and asymptotic communication capacity under given asynchronous levels are derived. On this basis, a two-stage optimization framework is developed to explicitly optimize the coordinated sensing waveform and communication beamforming, further enhancing the resistance to asynchrony. The framework leverages the Schur complement-based semi-definite relaxation to address the non-linear and non-convex challenges in the sensing design. For the communication stage, a conjugate gradient method with penalty terms is proposed to address the high complexity of the multi-variable non-convex problem. These efforts enhance coordinated sensing and communication in asynchronous scenarios, with simulation results provided to verify their effectiveness. © 1967-2012 IEEE.

关键词： Conjugate gradient method

来源：评论

学校读者我要写书评

暂无评论

Endogenous Security Through AI-Driven Physical-Layer Authentication for Future 6G Networks

引用

ZTE Communications 2025年第1期23卷 18-29页

作者： MENG Rui FAN Dayu XU Xiaodong LYU Suyu TAO Xiaofeng State Key Laboratory of Networking and Switching Technology Bei-jing University of Posts and TelecommunicationsBeijing 100876China Department of Broadband Communication Peng Cheng LaboratoryShenzhen 518066China School of Information Science and Technology Beijing University of TechnologyBeijing 100124China National Engineering Laboratory for Mobile Network Technologies Bei-jing University of Posts and TelecommunicationsBeijing 100876China

To ensure the access security of 6G,physical-layer authentication(PLA)leverages the randomness and space-time-frequency uniqueness of the channel to provide unique identity signatures for ***,the introduction of artificial intelligence(AI)facilitates the learning of the distribution characteristics of channel fingerprints,effectively addressing the uncertainties and unknown dynamic challenges in wireless link *** paper reviews representative AI-enabled PLA schemes and proposes a graph neural network(GNN)-based PLA approach in response to the challenges existing methods face in identifying mobile *** results demonstrate that the proposed method outperforms six baseline schemes in terms of authentication ***,this paper outlines the future development directions of PLA.

关键词： physical-layer authentication artificial intelligence wireless security intelligent authentication

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：