检索结果-内蒙古大学图书馆

ieee OPEN JOURNAL OF signal processing 2025年 6卷 303-311页

作者： Uliel, David Giryes, Raja Tel Aviv Univ Dept Elect Engn IL-6997801 Tel Aviv Israel

In unsupervised domain adaptation (UDA) labeled data is available for one domain (Source Domain) which is generated according to some distribution, and unlabeled data is available for a second domain (Target Domain) which is generated from a possibly different distribution but has the same task. The goal is to learn a model that performs well on the target domain although labels are available only for the source data. Many recent works attempt to align the source and the target domains by matching their marginal distributions in a learned feature space. In this paper, we address the domain difference as a nuisance, and enables better adaptability of the domains, by encouraging minimality of the target domain representation, disentanglement of the features, and a smoother feature space that cluster better the target data. To this end, we use the information bottleneck theory and a classical technique from the blind source separation framework, namely, ICA (independent components analysis). We show that these concepts can improve performance of leading domain adaptation methods on various domain adaptation benchmarks.

关键词： Feature extraction Training Loss measurement Data mining Random variables Independent component analysis Entropy Adaptation models Weight measurement Blind source separation domain adaptation information theory mutual information machine learning machine learning

来源：评论

学校读者我要写书评

暂无评论

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

引用

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

Complex Noise Suppression Using a Robust Dictionary learning Approach

引用

ieee GEOSCIENCE AND REMOTE SENSING LETTERS 2025年 22卷

作者： Feng, Zhenjie Anyang Inst Technol Sch Comp Sci & Informat Engn Anyang 455000 Peoples R China

Seismic data comprise reflection signals and various types of noise, including random high-frequency ambient noise, high-amplitude noise, and low-frequency ground-roll noise. Noise removal while protecting useful signals has been the long-term research focus of seismic exploration and data processing. Seldom methods can be used to deal with all these types of noise simultaneously, which motivates the study of this research. Here, we propose a systematic and effective method to tackle various kinds of complex seismic noise in seismic data by designing a dictionary learning approach. To deal with random and high-amplitude noise, we propose a robust inversion framework to estimate the signals from noise-corrupted data iteratively. The inversion scheme is inspired by an L1 -norm regularized iterative denoising framework but performs conveniently in terms of the coherency-promoting denoising operator. Then, we propose a bandpass-limited signal retrieving scheme internally within the iterative framework to tackle the ground-roll suppression issue. The iterative robust dictionary learning approach is based on a singular-value decomposition (SVD)-free approach to improve the efficiency without sacrificing the denoising performance. Results from synthetic and real data examples show promising performance of the proposed method.

关键词： Noise Noise reduction machine learning Dictionaries Vectors Iterative methods Geoscience and remote sensing Reflection Noise measurement Lubricating oils Dictionary learning ground-roll seismic data processing signal leakage

来源：评论

学校读者我要写书评

暂无评论

Low-Cost Driver Monitoring System Using Deep learning

引用

ieee ACCESS 2025年 13卷 14151-14164页

作者： Khalil, Hady A. Hammad, Sherif A. Abd El Munim, Hossam E. Maged, Shady A. Ain Shams Univ Fac Engn Dept Mechatron Engn Cairo 11571 Egypt Garraio Software Innovat Cairo 11816 Egypt Ain Shams Univ Fac Engn Dept Comp & Syst Engn Cairo 11571 Egypt

Driver monitoring systems are becoming an essential part of Advanced Driver Assistance Systems (ADAS) safety features in modern vehicles. The U.S. National Highway Traffic Safety Administration reports that drowsy/fatigued driving results in almost 100,000 road accidents per year. Driver's fatigue can have different causes, such as lack of sleep, long journeys, restlessness, mental pressure and alcohol consumption. Early monitoring systems relied on data from vehicle sensors, and modern systems commonly use driver's eye tracking. Recently, there has been growing interest in utilizing machine vision and deep learning for driver monitoring. Using machine vision can create more advanced driver monitoring systems capable of detecting driver attention state as well as other features like smartphone usage while driving and seat belts. machine vision systems usually require extensive processing power, which raises the cost of such systems. In this paper, we present a low-cost driver monitoring system using a 15 Raspberry Pi Zero 2 W board and deep learning CNN to deliver a system capable of monitoring and identifying different states of the driver like safe driving, distracted, drowsy, and smartphone usage, the system achieves an inference rate for 10 Frames Per Second (FPS) and above 90% accuracy with the testing dataset. In addition to the deep learning CNN which runs on Raspberry Pi CPU, we utilize the Raspberry Pi GPU to run a head pose estimation algorithm to boost the system's accuracy.

关键词： Vehicles Monitoring Computational modeling Accuracy Deep learning Quantization (signal) Cruise control Fatigue Training Pose estimation machine learning AI Raspberry Pi driver monitoring system tinyML YOLO OpenCL CNN embedded systems

来源：评论

学校读者我要写书评

暂无评论

Correlation-Boosted Ensemble Local Patterns for Photoplethysmographic signal Quality Classification

引用

ieee signal processing LETTERS 2025年 32卷 61-65页

作者： Lucafo, Giovani Lima, Rafael Sandoval, Italo Albany, Luz Penatti, Otavio Samsung R&D Inst Brazil BR-13097104 Campinas SP Brazil

Photoplethysmography (PPG) is a key component in a myriad of continuous and non-invasive health monitoring solutions, increasingly widespread in wearable devices, such as smartwatches and smart rings. Its high susceptibility to noise, such as motion artifacts and ambient light interference, however, can significantly hinder the learning process, as well as the resulting performance, of the deployed models. Given that, a signal Quality Assessment (SQA) auxiliary module is usually employed in such applications, for upfront selection of the PPG segments that should be used for reliable extraction of physiological information from the user. Most SQA strategies adopted in these devices rely either on (1) Deep learning (DL) models, capable of obtaining high performance metrics despite being of increased complexity and energy consumption, or (2) Decision Rule-based strategies that can assess signal quality in an increasedly energy-efficient way, albeit with reduced robustness. In this work, we introduce Hexa-SymmLTP-CC, a novel signal quality classifier composed of an ensemble of Local Pattern-based feature extractors followed by a downstream linear binary classifier, which outperforms state-of-the-art solutions, achieving accuracies of 93.93%, 96.06% and 96.55% across three clinical expert-annotated smartwatch and smart ring PPG datasets, while respecting the lightweight restrictions for wearable-based real-time monitoring applications.

关键词： Feature extraction Histograms Vectors Quality assessment Pipelines Analytical models Biomedical monitoring Visualization Standards Predictive models signal quality assessment machine learning local patterns photoplethysmography wearable

来源：评论

学校读者我要写书评

暂无评论

On the Impact of Model Compression for Bayesian Federated learning: An Analysis on Healthcare Data

引用

ieee signal processing LETTERS 2025年 32卷 251-255页

作者： Barbieri, Luca Savazzi, Stefano Nicoli, Monica Politecn Milan Dipartimento Ingn Gestionale I-20133 Milan Italy Inst Elect Informat Engn & Telecommun CNR I-20133 Milan Italy

Bayesian Federated learning (FL) policies enable multiple nodes to collaboratively train a shared machine learning (ML) model while accounting for the uncertainty of its predictions. This is accomplished by estimating the global posterior distribution in the model parameter space. Currently, Bayesian FL strategies are impaired by large communication costs that need to be reduced to provide more sustainable training platforms. This letter investigates the impact of compression strategies in centralized Bayesian FL setups, where a Parameter Sever (PS) is tasked to supervise the learning process. The goal is to study how compression affects the ability of Bayesian FL systems to provide high-quality, yet well-calibrated ML models. The analysis is carried out in the healthcare domain, where the prediction reliability is particularly critical, focusing on a medical imaging task. Numerical results show that applying aggressive compression policies highly reduces the ability of Bayesian FL systems to provide accurate and reliable ML models. On the contrary, light compression stages maximize accuracy and calibration at the cost of larger communication overheads.

关键词： Bayes methods Predictive models Data models Accuracy Training Medical services Vectors Uncertainty Costs Biomedical imaging Federated learning (FL) Bayesian deep learning healthcare networks Markov chain Monte Carlo compression

来源：评论

学校读者我要写书评

暂无评论

Obstacle Detection and Warning System for Visually Impaired Using IoT Sensors

引用

ieee ACCESS 2025年 13卷 35309-35321页

作者： Ikram, Sunnia Bajwa, Imran Sarwar Ikram, Amna Diez, Isabel de la Torre Rios, Carlos Eduardo Uc Castilla, Angel Kuc Islamia Univ Bahawalpur Dept Software Engn Bahawalpur 63100 Pakistan Islamia Univ Bahawalpur Dept Comp Sci Bahawalpur 63100 Pakistan Govt Sadiq Coll Women Univ Bahawalpur Dept Comp Sci Bahawalpur 63100 Pakistan Univ Valladolid Dept Signal Theory & Commun & Telemat Engn Campus Miguel Delibes Valladolid 47011 Spain Univ Europea Atlantico Project Management Res Grp Santander 00613 Spain Univ Int Iberoamer Dept Project Management Campeche 24560 Mexico Univ La Romana Dept Project Management La Romana 22000 Dominican Rep

Ensuring safe and independent mobility for visually impaired individuals requires efficient obstacle detection systems. This study introduces an innovative smart knee glove, integrating machine learning technologies for real-time obstacle detection and alerting. The system is equipped with ultrasonic sensor, PIR sensor and a buzzer, with data processing managed by an Arduino Uno microcontroller. To enhance detection accuracy, multiple machine learning algorithms including Decision Tree (DT), Support Vector machine (SVM), K-Nearest Neighbour (KNN), Random Forest (RF) and Gaussian Na & iuml;ve Bayes (GNB) are utilized. A novel Voting Classifier ensemble method is proposed, effectively combining the strengths of these classifiers to maximize performance. Rigorous cross-fold validation ensures robust evaluation under varying conditions. Experimental results demonstrates that the system achieves an impressive 98.34% detection accuracy within a 4-meter range, with high precision, recall and F1 scores. These findings underscore the system's reliability and potential to empower visually impaired users with safer, more autonomous navigation, marking a significant advancement in obstacle detection technologies.

关键词： Sensors Intelligent sensors Sensor systems Accuracy Feature extraction Navigation Real-time systems Acoustics Safety machine learning Obstacle detection IoT sensors visually impaired machine learning android application

来源：评论

学校读者我要写书评

暂无评论

Successive Refinement in Large-Scale Computation: Expediting Model Inference Applications

引用

ieee TRANSACTIONS ON signal processing 2025年 73卷 811-826页

作者： Esfahanizadeh, Homa Cohen, Alejandro Shamai, Shlomo Medard, Muriel Nokia Bell Labs Murray Hill NJ 07974 USA Technion Israel Inst Technol Elect & Comp Engn Dept IL-3200003 Haifa Israel MIT Res Lab Elect RLE Cambridge MA 02139 USA

Modern computationally-intensive applications often operate under time constraints, necessitating acceleration methods and distribution of computational workloads across multiple entities. However, the outcome is either achieved within the desired timeline or not, and in the latter case, valuable resources are wasted. In this paper, we introduce solutions for layered-resolution computation. These solutions allow lower-resolution results to be obtained at an earlier stage than the final result. This innovation notably enhances the deadline-based systems, as if a computational job is terminated due to time constraints, an approximate version of the final result can still be generated. Moreover, in certain operational regimes, a high-resolution result might be unnecessary, because the low-resolution result may already deviate significantly from the decision threshold, for example in AI-based decision-making systems. Therefore, operators can decide whether higher resolution is needed or not based on intermediate results, enabling computations with adaptive resolution. We present our framework for two critical and computationally demanding jobs: distributed matrix multiplication (linear) and model inference in machine learning (nonlinear). Our theoretical and empirical results demonstrate that the execution delay for the first resolution is significantly shorter than that for the final resolution, while maintaining overall complexity comparable to the conventional one-shot approach. Our experiments further illustrate how the layering feature increases the likelihood of meeting deadlines and enables adaptability and transparency in massive, large-scale computations.

关键词： Computation layered resolution adaptive linear nonlinear machine learning inference matrix multiplication Computation layered resolution adaptive linear nonlinear machine learning inference matrix multiplication

来源：评论

学校读者我要写书评

暂无评论

Weighted Average Consensus Algorithms in Distributed and Federated learning

引用

ieee TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING 2025年第2期12卷 1369-1382页

作者： Tedeschini, Bernardo Camajori Savazzi, Stefano Nicoli, Monica Politecn Milan I-20133 Milan Italy Consiglio Nazl Ric CNR I-20133 Milan Italy

The exponential growth of the Internet of Things (IoT) has created an essential demand for Distributed machine learning (DML) systems. In this context, Federated learning (FL) allows IoT devices to collaboratively train models while maintaining data ownership and privacy. Despite the evident advantages, FL faces practical challenges such as client selection and adaptation to heterogeneous data distributions. Recently, consensus-driven algorithms have been proposed to enable efficient and scalable FL without a central coordinating entity. Weighted Average Consensus (WAC) tools, primarily used in distributed signal processing, fail to address FL-specific challenges. The paper proposes a new family of server-less FL algorithms optimized to exploit WAC techniques. In particular, we propose an evolution of the centralized Federated Adaptive Weighting (FedAdp) method and present three distinct WAC schemes specifically designed for non-Independent and Identical Distributed (IID) data. Each scheme has a unique aggregation part that optimizes the weights of the clients' local models. The performances are evaluated in a real-world IoT system, analyzing their convergence properties in the context of heterogeneous client populations. Results show that the proposed algorithms outperform vanilla consensus FL up to 56% of accuracy and they are resilient to both label and sample data skewness.

关键词： Data models Heuristic algorithms Internet of Things Adaptation models Convergence signal processing algorithms Vehicle dynamics Optimization Distributed databases Training FL over networks weighted average consensus non-independent and identical distributed edge computing

来源：评论

学校读者我要写书评

暂无评论

Representation Transfer learning via Multiple Pre-Trained Models for Linear Regression

引用

ieee JOURNAL OF SELECTED TOPICS IN signal processing 2025年第1期19卷 208-220页

作者： Singh, Navjot Diggavi, Suhas Univ Calif Los Angeles Dept Elect & Comp Engn Los Angeles CA 90025 USA

In this paper, we consider the problem of learning a linear regression model on a data domain of interest (target) given few samples. To aid learning, we are provided with a set of pre-trained regression models that are trained on potentially different data domains (sources). Assuming a representation structure for the data generating linear models at the sources and the target domains, we propose a representation transfer based learning method for constructing the target model. The proposed scheme is comprised of two phases: (i) utilizing the different source representations to construct a representation that is adapted to the target data, and (ii) using the obtained model as an initialization to a fine-tuning procedure that re-trains the entire (over-parameterized) regression model on the target data. For each phase of the training method, we provide excess risk bounds for the learned model compared to the true data generating target model. The derived bounds show a gain in sample complexity for our proposed method compared to the baseline method of not leveraging source representations when achieving the same excess risk, therefore, theoretically demonstrating the effectiveness of transfer learning for linear regression.

关键词： Data models Adaptation models Training Transfer learning Vectors Linear regression Analytical models Few shot learning Context modeling Complexity theory machine learing: Transfer learning few shot learning representation learning

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：