检索结果-内蒙古大学图书馆

Client Selection for Generalization in Accelerated Federated Learning: A Multi-Armed Bandit Approach

IEEE ACCESS 2025年 13卷 33697-33713页

作者： Ben Ami, Dan Cohen, Kobi Zhao, Qing Ben Gurion Univ Negev Sch Elect & Comp Engn IL-8410501 Beer Sheva Israel Cornell Univ Dept Elect & Comp Engn Ithaca NY 14853 USA

Federated learning (FL) is an emerging machine learning (ML) paradigm used to train models across multiple nodes (i.e., clients) holding local data sets, without explicitly exchanging the data. It has attracted a growing interest in recent years due to its advantages in terms of privacy considerations, and communication resources. In FL, selected clients train their local models and send a function of the models to the server, which consumes a random processing and transmission time. The server updates the global model and broadcasts it back to the clients. The client selection problem in FL is to schedule a subset of the clients for training and transmission at each given time so as to optimize the learning performance. In this paper, we present a novel multi-armed bandit (MAB)-based approach for client selection to minimize the training latency without harming the ability of the model to generalize, that is, to provide reliable predictions for new observations. We develop a novel algorithm to achieve this goal, dubbed Bandit Scheduling for FL (BSFL). We analyze BSFL theoretically, and show that it achieves a logarithmic regret, defined as the loss of BSFL as compared to a genie that has complete knowledge about the latency means of all clients. We conducted evaluations under both i.i.d. and non-i.i.d. scenarios using a synthetic dataset with a linear regression model and two well-known datasets, Fashion-MNIST and CIFAR-10 with CNN-based classification models. The results demonstrate that BSFL outperforms existing methods.

关键词： Training Servers Optimization Hands Federated learning Data models signal processing algorithms Simulated annealing History Standards Federated learning (FL) client selection client scheduling multi-armed bandit (MAB) generalization in machine learning

来源：评论

学校读者我要写书评

暂无评论

Polynomial Graphical Lasso: Learning Edges From Gaussian Graph-Stationary signals

引用

IEEE TRANSACTIONS ON signal processing 2025年 73卷 1153-1167页

作者： Buciulea, Andrei Ying, Jiaxi Marques, Antonio G. Palomar, Daniel P. King Juan Carlos Univ Dept Signal Theory & Commun Madrid 28933 Spain Hong Kong Univ Sci & Technol Dept Math Kowloon Clear Water Bay Hong Kong Peoples R China Hong Kong Univ Sci & Technol Dept Elect & Comp Engn Kowloon Clear Water Bay Hong Kong Peoples R China Hong Kong Univ Sci & Technol Dept Ind Engn & Decis Analyt Kowloon Clear Water Bay Hong Kong Peoples R China

This paper introduces Polynomial Graphical Lasso (PGL), a new approach to learning graph structures from nodal signals. Our key contribution lies in modeling the signals as Gaussian and stationary on the graph, enabling the development of a graph learning formulation that combines the strengths of graphical lasso with a more encompassing model. Specifically, we assume that the precision matrix can take any polynomial form of the sought graph, allowing for increased flexibility in modeling nodal relationships. Given the inherent complexity and nonconvexity of the optimization problem, we (i) propose a low-complexity algorithm that alternates between estimating the graph and precision matrices, and (ii) characterize its convergence. We evaluate the performance of PGL through comprehensive numerical simulations using both synthetic and real data, demonstrating its superiority over several alternatives. Overall, this approach presents a significant advancement in graph learning and holds promise for various applications in graph-aware signal analysis and beyond.

关键词： Polynomials Covariance matrices Sparse matrices Topology Optimization Filters Convergence signal processing algorithms Correlation Vectors Network-topology inference graph learning Gaussian signals graph-stationary signals covariance estimation

来源：评论

学校读者我要写书评

暂无评论

Practical Compact Deep Compressed Sensing

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2025年第3期47卷 1610-1626页

作者： Chen, Bin Zhang, Jian Peking Univ Sch Elect & Comp Engn Shenzhen 518055 Peoples R China

Recent years have witnessed the success of deep networks in compressed sensing (CS), which allows for a significant reduction in sampling cost and has gained growing attention since its inception. In this paper, we propose a new practical and compact network dubbed PCNet for general image CS. Specifically, in PCNet, a novel collaborative sampling operator is designed, which consists of a deep conditional filtering step and a dual-branch fast sampling step. The former learns an implicit representation of a linear transformation matrix into a few convolutions and first performs adaptive local filtering on the input image, while the latter then uses a discrete cosine transform and a scrambled block- diagonal Gaussian matrix to generate under-sampled measurements. Our PCNet is equipped with an enhanced proximal gradient descent algorithm-unrolled network for reconstruction. It offers flexibility, interpretability, and strong recovery performance for arbitrary sampling rates once trained. Additionally, we provide a deployment-oriented extraction scheme for single-pixel CS imaging systems, which allows for the convenient conversion of any linear sampling operator to its matrix form to be loaded onto hardware like digital micro-mirror devices. Extensive experiments on natural image CS, quantized CS, and self-supervised CS demonstrate the superior reconstruction accuracy and generalization ability of PCNet compared to existing state-of-the-art methods, particularly for high-resolution images.

关键词： Artificial neural networks Image reconstruction Imaging Matrix converters Hardware Filtering Training signal processing algorithms Optimization Micromirrors Compressed sensing sampling matrix single-pixel imaging algorithm unrolling and structural reparameterization

来源：评论

学校读者我要写书评

暂无评论

Damped Sine Wave Parameter Extraction: Application to QCM-D signals for Accurate Measurements

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2025年 74卷

作者： Fort, Ada Landi, Elia Moretti, Riccardo Carbone, Paolo Moschitta, Antonio Univ Siena Dept Informat Engn & Math I-53100 Siena Italy Univ Perugia Dept Engn I-06125 Perugia Italy

Quartz crystal microbalance with dissipation (QCM-D) monitoring is a sensing technique that finds diverse applications in chemical sensing, biosensing, and industrial fields. By monitoring the vibrational characteristics of a thin piezoelectric SiO2 resonator, this technique provides valuable information about the viscoelastic behavior of media in contact with its surface. This article evaluates the performance of algorithms applied to 10 MHz AT-cut QCM-D signals for estimating the frequency and time constant related to the fundamental resonant mode, extending the study to assess the effectiveness of disturbance rejection, particularly against quartz spurious oscillating modes. To this end, both damped single-tone and three-tone signals are analyzed to cover all potential outputs of a QCM-D in its fundamental bulk shear mode. Various estimation algorithms are considered: one operating in the frequency domain and others in the time domain, based on model fitting techniques with novel preprocessing steps for optimized initialization. The analysis utilizes signals from a QCM-D measurement set-up and signals simulated according to the characteristics of the AT-CUT crystals used. Among model fitting-based estimators, two techniques were developed for single-tone and three-tone signals, splitting the model-fitting problem into nonlinear and linear subproblems, demonstrating the best accuracy and processing time. The developed techniques meet strict measurement objectives, achieving frequency measurement resolution around 1 ppm (10 Hz) and decay time accuracy around 1 mu s, even with additive white Gaussian noise (AWGN) standard deviations of about 1% of the fundamental tone amplitude in water. These techniques are applicable in various contexts where recovering parameters from damped sine waves in the presence of white noise is required.

关键词： Resonant frequency Resonance Frequency measurement Transient analysis Q-factor Impedance Time-frequency analysis Quartz crystals Fitting Accuracy Quartz crystal microbalance with dissipation (QCM-D) monitoring signal processing algorithms

来源：评论

学校读者我要写书评

暂无评论

Research on the Application of Deep Learning-Based Image Recognition Algorithm in signal processing

Research on the Application of Deep Learning-Based Image Rec...

引用

Power Electronics, Computer Applications (ICPECA), IEEE International Conference on

作者： Jing Lv Hunan Chemical Vocational and Technical College Zhuzhou China

ISBN: (数字)9798331533694

ISBN: (纸本)9798331533700

A GAN-based image recognition algorithm is presented to solve these problems. Firstly, the GAN frame is composed of a generator and a discriminator. The generator can produce real images or remove noise by learning the signal image. Then, the image is compared with the real image. In the pre-processing phase, the standard, denotation and enhancement of the signal image are carried out to guarantee the data quality and the validity of the characteristic. Experiments show that the precision of GAN-based image recognition is up to 96.3%, which is more than 15% than traditional methods (e.g., SVM). In the signal denoising task, the model significantly improves the signal-to-noise ratio (SNR) by 20% and effectively removes image noise. In addition, GAN also performs extremely well in pattern recognition tasks, with a correct recognition rate of 94.7%. Compared with traditional deep learning algorithms, the GAN model shows stronger robustness when processing signals with more noise or incomplete data.

关键词： Deep learning Training Image recognition signal processing algorithms Generative adversarial networks Generators Robustness Pattern recognition Image denoising signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Efficient signal processing: Harnessing Distributed Arithmetic for High-Speed FFT Operations

Efficient Signal Processing: Harnessing Distributed Arithmet...

引用

Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE), International Conference

作者： M. Bharathi Krithikaa Mohanarangam Yasha Jyothi M Shirur ECE Mohan Babu University (Erstwhile Sree Vidyanikethan Engineering College) Tirupati Andhra Pradesh India Symbiosis Institute of Technology Symbiosis International (Deemed University) Pune India Electronics and Communication Engineering BNM Institute of Technology Bangalore India

ISBN: (数字)9798331515911

ISBN: (纸本)9798331515928

The present article introduces a novel approach to significantly enhance the computational efficiency of modern Digital signal processing (DSP) processors. The algorithm enhances the efficiency of Sum of Products (SOP) operations, which are essential for real-time signal processing applications, by utilizing Distributed Arithmetic (DA) techniques for the Fast Fourier Transform (FFT) and its inverse (Inverse FFT) (IDFT) computations. The present study undertakes an exhaustive quantitative analysis to assess the computational delays of the proposed 2, 4, and 8-point Fast Fourier Transform (FFT) and Inverse Discrete Fourier Transform (IDFT) across various bit representations (4, 8, and 16-bit). For example, when using 4-bit representations, the computational delays for a 2-point Fast Fourier Transform (FFT) are measured to be 13.318ns, 30.629ns for a 4-point FFT, and 27.779ns for an 8-point FFT. Similarly, the delays for the 2-point Fast Fourier Transform (FFT) in an 8-bit representation are 13.727ns, the 4-point FFT in 22.276ns, and the 8-point FFT in 30.592ns. The study further includes the addition of appropriate delays for IDFT operations, which enhances the understanding of the effectiveness of the approach suggested in various scenarios. Furthermore, the research presents and evaluates three distinct categories of transforms, namely 2-point, 4-point, and 8-point FFT and IDFT structures. These structures utilize different quantities of Look-Up Tables (LUTs) for computational purposes. The research showcases the exceptional computing efficiency of the suggested design through detailed quantitative analysis, emphasizing its potential advantages in improving DSP implementations in practical situations.

关键词： Fast Fourier transforms Statistical analysis signal processing algorithms Digital signal processing Real-time systems Delays Table lookup Computational efficiency Velocity measurement Arithmetic

来源：评论

学校读者我要写书评

暂无评论

Two-Stage Spatial Whitening and Normalized MUSIC for Robust DOA Estimation of GNSS signals Under Jamming

引用

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS 2025年第2期61卷 2557-2572页

作者： Wang, Chuanrui Cui, Xiaowei Liu, Gang Lu, Mingquan Tsinghua Univ Dept Elect Engn Beijing 100190 Peoples R China Tsinghua Univ Beijing Natl Res Ctr Informat Sci & Technol Beijing 100190 Peoples R China

The estimation of the direction of arrival (DOA) of navigation signals is a critical function of a global navigation satellite system (GNSS) array receiver for applications such as attitude determination, spoofing detection, and blind adaptive beamforming. However, high-power GNSS jamming poses significant challenges to accurate DOA estimation. This paper presents spatial whitening and normalized multiple signal classification (SWN-MUSIC), an enhanced two-stage MUSIC algorithm for GNSS signal DOA estimation under jamming. In the first stage, inspired by minimum variance distortionless response (MVDR) beamforming, we use a spatial whitening matrix for matrix filtering to achieve jamming suppression. In the second stage, an approach accounts for matrix filtering effects when computing the MUSIC pseudospectrum. Compared to prior GNSS DOA methods under jamming, our SWN-MUSIC shows improved accuracy, particularly when the DOA of jamming and signals are closely aligned. Additionally, we develop a reduced-dimension SWN-MUSIC (RD-SWN-MUSIC) algorithm to lower the complexity of applying SWN-MUSIC for two-dimensional DOA estimation using uniform rectangular arrays (URA) under jamming. Compared to standard reduced-dimension MUSIC, our RD-SWN-MUSIC avoids unnecessary precision loss by performing separate one-dimensional parameter searches for each DOA parameter. Numerical simulations indicate that RD-SWN-MUSIC closely approaches the Cramer-Rao lower bound (CRLB), with significantly lower complexity than classical MUSIC and its previously proposed versions.

关键词： Jamming Direction-of-arrival estimation Global navigation satellite system Estimation Multiple signal classification Array signal processing Antenna arrays Vectors Transmission line matrix methods signal processing algorithms Antenna array direction of arrival (direction of arrival) global navigation satellite system (GNSS) jamming multiple signal classification (MUSIC) algorithm

来源：评论

学校读者我要写书评

暂无评论

Data-Driven Distributed Spectrum Estimation for Linear Time-Invariant Systems

引用

IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS 2025年第1期12卷 1125-1136页

作者： Liu, Shenyu Cortes, Jorge Martinez, Sonia Beijing Inst Technol Sch Automat Beijing 100081 Peoples R China Univ Calif San Diego Dept Aerosp & Mech Engn San Diego CA 92093 USA

This article tackles spectrum estimation of a linear time-invariant system by a multiagent network using data. We consider a group of agents that communicate over a strongly connected, aperiodic graph and do not have any knowledge of the system dynamics. Each agent only measures some signals that are linear functions of the system states or inputs, and does not know the functional form of this dependence. The proposed distributed algorithm consists of two steps that rely on the collected data: first, the identification of an unforced trajectory of the system, and second, the estimation of the coefficients of the characteristic polynomial of the system matrix using this unforced trajectory. We show that each step can be formulated as a problem of finding a common solution to a set of linear algebraic equations, which are amenable to distributed algorithmic solutions. We prove that under mild assumptions on the collected data, when the initial condition of the system is random, the proposed distributed algorithm accurately estimates the spectrum with probability 1.

关键词： Spectral analysis Distributed algorithms Linear systems Estimation System identification signal processing algorithms Trajectory Data-driven modeling discrete-time systems distributed processing system identification time invariant systems

来源：评论

学校读者我要写书评

暂无评论

A Mirror Descent-Based Algorithm for Corruption-Tolerant Distributed Gradient Descent

引用

IEEE TRANSACTIONS ON signal processing 2025年 73卷 827-842页

作者： Wang, Shuche Tan, Vincent Y. F. Natl Univ Singapore Inst Operat Res & Analyt Singapore 117602 Singapore Natl Univ Singapore Dept Math Dept Elect & Comp Engn Singapore 117602 Singapore Natl Univ Singapore Inst Operat Res & Analyt Singapore 117602 Singapore

Distributed gradient descent algorithms have come to the fore in modern machine learning, especially in parallelizing the handling of large datasets that are distributed across several workers. However, scant attention has been paid to analyzing the behavior of distributed gradient descent algorithms in the presence of adversarial corruptions instead of random noise. In this paper, we formulate a novel problem in which adversarial corruptions are present in a distributed learning system. We show how to use ideas from (lazy) mirror descent to design a corruption-tolerant distributed optimization algorithm. Extensive convergence analysis for (strongly) convex loss functions is provided for different choices of the stepsize. We carefully optimize the stepsize schedule to accelerate the convergence of the algorithm, while at the same time amortizing the effect of the corruption over time. Experiments based on linear regression, support vector classification, and softmax classification on the MNIST dataset corroborate our theoretical findings.

关键词： Vectors Servers Noise measurement signal processing algorithms Machine learning algorithms Convergence Uplink Mirrors Downlink Machine learning Distributed gradient descent mirror descent adversarial corruptions noisy channels convergence rates

来源：评论

学校读者我要写书评

暂无评论

Direction of Arrival Estimation Based on AR Modeling in the Presence of Colored Noise

引用

IEEE ACCESS 2025年 13卷 62427-62433页

作者： Bakhshandeh Babarsad, Sina Mahmoudi, Alimorad Namjoo, Ehsan Shahid Chamran Univ Ahvaz Fac Engn Dept Elect Engn Ahvaz *** Iran

This paper investigates the problem of estimating the direction of arrival considering colored noise at the sensors. To this aim, we model the recieved signal at each time instant with an AR system with rank equal to the number of sources and we simultaneously estimated signal and noise parameters by solving a quadratic eigenvalue problem. Simulation results demonstrate that utilizing root mean square error as the accuracy metric, the proposed method outperforms existing subspace-based approaches, particularly when the targets are close to eachother.

关键词： Noise Direction-of-arrival estimation Sensors Colored noise Maximum likelihood estimation Transforms signal processing algorithms Multiple signal classification Eigenvalues and eigenfunctions White noise AR modeling colored noise DOA estimation subspace-based methods

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：