检索结果-内蒙古大学图书馆

Federated Generalized Bayesian Learning via Distributed Stein Variational Gradient Descent

IEEE TRANSACTIONS ON signal processing 2022年 70卷 2180-2192页

作者： Kassab, Rahif Simeone, Osvaldo Kings Coll London Kings Commun Learning & Informat Proc KCLIP Lab London WC2B 4BG England

This paper introduces Distributed Stein Variational Gradient Descent (DSVGD), a non-parametric generalized Bayesian inference framework for federated learning. DSVGD maintains a number of non-random and interacting particles at a central server to represent the current iterate of the model global posterior. The particles are iteratively downloaded and updated by a subset of agents with the end goal of minimizing the global free energy. By varying the number of particles, DSVGD enables a flexible trade-off between per-iteration communication load and number of communication rounds. DSVGD is shown to compare favorably to benchmark frequentist and Bayesian federated learning strategies in terms of accuracy and scalability with respect to the number of agents, while also providing well-calibrated, and hence trustworthy, predictions.

关键词： Bayes methods Servers Collaborative work Uncertainty signal processing algorithms Convergence Training Federated learning bayesian learning variational inference

来源：评论

学校读者我要写书评

暂无评论

A Novel Adaptive Scheme to Improve the Performance of Feedforward Active Vibration Control Systems

引用

IEEE-ASME TRANSACTIONS ON MECHATRONICS 2022年第4期27卷 2322-2332页

作者： Ortiz Batista, Eduardo Luiz Barghouthi, Maurizio Radloff de Oliveira Lopes, Eduardo Marcio Univ Fed Santa Catarina Circuits & Signal Proc Lab Dept Elect & Elect Engn BR-88040900 Florianopolis SC Brazil Univ Fed Parana Dept Mech Engn Res Grp Vibrat & Sound Mech Syst BR-81530900 Curitiba Parana Brazil

Feedforward active vibration control techniques aim at canceling vibrations occurring in a system by means of introducing intentional forces that generate destructively interfering vibrations. These forces are typically controlled using an adaptive filtering algorithm that tracks the system state in real time and seeks to minimize overall vibration. One important often-overlooked issue in such systems is related with the vibrations introduced by the adaptive control algorithm itself either during its transient learning phase or due to its nonideal characteristic. These vibrations may become very significant in several mechanical structures due to resonance phenomena. In this context, this work is dedicated to the development of an effective scheme for feedforward active vibration control that is capable of dealing with undesirable vibrations introduced by the adaptive controller. A particular version of the filtered-x normalized least-mean-squares algorithm tailored to the proposed scheme is also developed. Results obtained from numerical simulation as well as from an experimental device are presented, aiming to demonstrate the effectiveness of the proposed scheme.

关键词： Vibrations Vibration control Feedforward systems Structural beams signal processing algorithms Real-time systems Force Adaptive filter filtered-x normalized least-mean-squares (FxNLMS) algorithm vibration control

来源：评论

学校读者我要写书评

暂无评论

Covariance-Free Sparse Bayesian Learning

引用

IEEE TRANSACTIONS ON signal processing 2022年 70卷 3818-3831页

作者： Lin, Alexander Song, Andrew H. Bilgic, Berkin Ba, Demba Harvard Univ Sch Engn & Appl Sci Cambridge MA 02138 USA MIT Elect Engn & Comp Sci Dept Cambridge MA 02138 USA MIT Harvard MIT Hlth Sci & Technol Cambridge MA 02138 USA Athinoula A Martinos Ctr Biomed Imaging Charlestown MA 02129 USA Harvard Med Sch Dept Radiol Boston MA 02115 USA

Sparse Bayesian learning (SBL) is a powerful framework for tackling the sparse coding problem while also providing uncertainty quantification. The most popular inference algorithms for SBL exhibit prohibitively large computational costs for high-dimensional problems due to the need to maintain a large covariance matrix. To resolve this issue, we introduce a new method for accelerating SBL inference - named covariance-free expectation maximization (CoFEM) - that avoids explicit computation of the covariance matrix. CoFEM solves multiple linear systems to obtain unbiased estimates of the posterior statistics needed by SBL. This is accomplished by exploiting innovations from numerical linear algebra such as preconditioned conjugate gradient and a little-known diagonal estimation rule. For a large class of compressed sensing matrices, we provide theoretical justifications for why our method scales well in high-dimensional settings. Through simulations, we show that CoFEM can be up to thousands of times faster than existing baselines without sacrificing coding accuracy. Through applications to calcium imaging deconvolution and multi-contrast MRI reconstruction, we show that CoFEM enables SBL to tractably tackle high-dimensional sparse coding problems of practical interest.

关键词： signal processing algorithms Inference algorithms Bayes methods Encoding Covariance matrices Linear systems Computational modeling Sparse Bayesian learning (SBL) compressed sensing expectation-maximization algorithms computational efficiency

来源：评论

学校读者我要写书评

暂无评论

Explainability in Graph Data Science Interpretability, replicability, and reproducibility of community detection

引用

IEEE signal processing MAGAZINE 2022年第4期39卷 25-39页

作者： Aviyente, Selin Karaaslanli, Abdullah Michigan State Univ Dept Elect & Comp Engn E Lansing MI 48824 USA Michigan State Univ Elect & Comp Engn E Lansing MI 48824 USA

In many modern data science problems, data are represented by a graph (network), e.g., social, biological, and communication networks. Over the past decade, numerous signal processing and machine learning (ML) algorithms have been introduced for analyzing graph structured data. With the growth of interest in graphs and graph-based learning tasks in a variety of applications, there is a need to explore explainability in graph data science. In this article, we aim to approach the issue of explainable graph data science, focusing on one of the most fundamental learning tasks, community detection, as it is usually the first step in extracting information from graphs. A community is a dense subnetwork within a larger network that corresponds to a specific function. Despite the success of different community detection methods on synthetic networks with strong modular structure, much remains unknown about the quality and significance of the outputs of these algorithms when applied to real-world networks with unknown modular structure. Inspired by recent advances in explainable artificial intelligence (AI) and ML, in this article, we present methods and metrics from network science to quantify three different aspects of explainability, i.e., interpretability, replicability, and reproducibility, in the context of community detection.

关键词： Machine learning algorithms signal processing algorithms Machine learning Data science signal processing Reproducibility of results

来源：评论

学校读者我要写书评

暂无评论

A Steering-Matrix-Based Multiobjective Evolutionary Algorithm for High-Dimensional Feature Selection

引用

IEEE TRANSACTIONS ON CYBERNETICS 2022年第9期52卷 9695-9708页

作者： Cheng, Fan Chu, Feixiang Xu, Yi Zhang, Lei Anhui Univ Sch Comp Sci & Technol Key Lab Intelligent Comp & Signal Proc Minist Educ Hefei 230039 Peoples R China Shanghai Jiao Tong Univ Sch Comp Sci & Technol Shanghai 200139 Peoples R China

In recent years, multiobjective evolutionary algorithms (MOEAs) have been demonstrated to show promising performance in feature selection (FS) tasks. However, designing an MOEA for high-dimensional FS is more challenging due to the curse of dimensionality. To address this problem, in this article, a steering-matrix-based multiobjective evolutionary algorithm, called SM-MOEA, is proposed. In SM-MOEA, a steering matrix is suggested and harnessed to guide the evolution of the population, which not only improves the search efficiency greatly but also obtains the feature subsets with high quality. Specifically, each element SM(i, j) in the steering matrix SM reflects the probability of the jth feature that is selected in the ith individual (feature subset), which is generated by considering the importance of both the feature j and the individual i. Based on the suggested steering matrix, two important operators referred to as dimensionality reduction and individual repairing operators are developed to effectively steer the population evolution in each generation. In addition, an effective initialization and update strategy for the steering matrix is also designed to further improve the performance of SM-MOEA. The experimental results on 12 high-dimensional datasets with the number of features ranging from 3000 to 13,000 demonstrate the superiority of the proposed algorithm over several state-of-the-art algorithms (including single-objective and MOEAs for high-dimensional FS) in terms of both the number and quality of the selected features.

关键词： Optimization signal processing algorithms Sociology Feature extraction Task analysis Support vector machines Error analysis Evolutionary algorithm feature selection (FS) high-dimensional data multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Deep Unfolding With Normalizing Flow Priors for Inverse Problems

引用

IEEE TRANSACTIONS ON signal processing 2022年 70卷 2962-2971页

作者： Wei, Xinyi van Gorp, Hans Gonzalez-Carabarin, Lizeth Freedman, Daniel Eldar, Yonina C. van Sloun, Ruud J. G. Eindhoven Univ Technol Dept Elect Engn NL-5612 AZ Eindhoven Netherlands Google Life Sci Verily Res IL-3190500 Haifa Israel Weizmann Inst Sci Dept Math & Comp Sci IL-7610001 Rehovot Israel

Many application domains, spanning from computational photography to medical imaging, require recovery of high-fidelity images from noisy, incomplete or partial/compressed measurements. State-of-the-art methods for solving these inverse problems combine deep learning with iterative model-based solvers, a concept known as deep algorithm unfolding or unrolling. By combining a-priori knowledge of the forward measurement model with learned proximal image-to-image mappings based on deep networks, these methods yield solutions that are both physically feasible (data-consistent) and perceptually plausible (consistent with prior belief). However, current proximal mappings based on (predominantly convolutional) neural networks only implicitly learn such image priors. In this paper, we propose to make these image priors fully explicit by embedding deep generative models in the form of normalizing flows within the unfolded proximal gradient algorithm, and training the entire algorithm end-to-end for a given task. We demonstrate that the proposed method outperforms competitive baselines on various image recovery tasks, spanning from image denoising to inpainting and deblurring, effectively adapting the prior to the restoration task at hand.

关键词： Task analysis signal processing algorithms Optimization Image reconstruction Deep learning Standards Noise measurement Deep unfolding normalizing flows inverse problem image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Plug-and-Play Regularization Using Linear Solvers

引用

IEEE TRANSACTIONS ON IMAGE processing 2022年 31卷 6344-6355页

作者： Nair, Pravin Chaudhury, Kunal N. Indian Inst Sci Dept Elect Engn Bengaluru 560012 India

There has been tremendous research on the design of image regularizers over the years, from simple Tikhonov and Laplacian to sophisticated sparsity and CNN-based regularizers. Coupled with a model-based loss function, these are typically used for image reconstruction within an optimization framework. The technical challenge is to develop a regularizer that can accurately model realistic images and be optimized efficiently along with the loss function. Motivated by the recent plug-and-play paradigm for image regularization, we construct a quadratic regularizer whose reconstruction capability is competitive with state-of-the-art regularizers. The novelty of the regularizer is that, unlike classical regularizers, the quadratic objective function is derived from the observed data. Since the regularizer is quadratic, we can reduce the optimization to solving a linear system for applications such as superresolution, deblurring, inpainting, etc. In particular, we show that using iterative Krylov solvers, we can converge to the solution in few iterations, where each iteration requires an application of the forward operator and a linear denoiser. The surprising finding is that we can get close to deep learning methods in terms of reconstruction quality. To the best of our knowledge, the possibility of achieving near state-of-the-art performance using a linear solver is novel.

关键词： Kernel Image reconstruction Convergence Superresolution signal processing algorithms Optimization Linear systems Image reconstruction regularization Gaussian denoiser plug-and-play method Krylov solver

来源：评论

学校读者我要写书评

暂无评论

Recognition of Human Activity using signal processing & Deep Neural Networks

Recognition of Human Activity using Signal Processing & Deep...

引用

International Conference for Convergence of Technology (I2CT)

作者： Bhimasen Moharana Bhramara Bar Biswal Tapaswini Samant Trupti Mayee Behera Subhashree Mishra Shobhan Banerjee Department of Computer Science & Engineering Lovely Professional University Punjab India P.G. Department of Computer Science & Applications College of Engineering Bhubaneswar India School of Electronics Engineering Kalinga Institute of Industrial Technology Bhubaneswar India Department of Engineering Technology Birla Institute of Technology and Science - Pilani Rajasthan India

The task of tracking the activities being performed by humans is of immense importance as it finds application in healthcare, marketing, surveillance, etc. To keep track of this, they need to be recognized and classified into various activities. The data can be collected through smart devices such as smartwatches or smartphones which internally use various sensors to measure the physical movements of the person using it. We have used Human Activity Recognition Dataset from the UCI ML Repository, which consists of the data being collected from the smartphones of 30 volunteers. After choosing the appropriate features, we trained an Artificial Neural Network and a Long Short Term Memory Neural Network to classify the activities into various categories and performed a comparative analysis of these neural networks among themselves & also with respect to the previously leveraged techniques & algorithms on the dataset.

关键词： Performance evaluation Tracking Surveillance signal processing algorithms Medical services signal processing Task analysis

来源：评论

学校读者我要写书评

暂无评论

Algorithm Research on the Conductor Eccentricity of a Circular Dot Matrix Hall High Current Sensor for ITER

引用

IEEE TRANSACTIONS ON PLASMA SCIENCE 2022年第6期50卷 1962-1970页

作者： Wu, Xu Huang, Haihong Peng, Lan Huang, Ya Wang, Yuhang Hefei Univ Technol Sch Elect Engn & Automat Hefei 230009 Peoples R China Chinese Acad Sci Inst Plasma Phys Hefei Inst Phys Sci Hefei 230031 Peoples R China

To meet the high current measurement requirements of the international thermonuclear experimental reactor polar field (ITER PF) power supply, a noncontact dot matrix Hall high current sensor without a magnetic core is designed. Due to the existence of installation error, the measurement accuracy will be reduced due to conductor eccentricity in practical use. To solve this problem, this article will take the circular dot matrix Hall current sensor composed of eight Hall elements as an example and note the shortcomings of the two algorithms by analyzing the numerical integration algorithm and conductor positioning algorithm. Based on the two algorithms, a digital signal processing (DSP) conditioning algorithm is proposed. The DSP conditioning algorithm detects the output voltage of each Hall element after conductor offset and adjusts the output voltage of each Hall element by self-feedback, which simplifies the complexity of algorithm operation and eliminates conductor eccentricity error at the same time. Through simulation and field measurement verification, DSP self-feedback conditioning of the Hall element output voltage can eliminate the influence of conductor eccentricity error and improve the measurement accuracy of the sensor.

关键词： Conductors Current measurement signal processing algorithms Measurement uncertainty Heuristic algorithms Temperature measurement Sensitivity Circular dot matrix Hall sensor conductor eccentricity conductor positioning digital signal processing (DSP) conditioning numerical integration

来源：评论

学校读者我要写书评

暂无评论

Analysis of the Kalman Filter with the MPU6050 Accelerometer and Gyroscope 15

Analysis of the Kalman Filter with the MPU6050 Accelerometer...

引用

15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024

作者： Kumar, Pranav Swarup Vignesh, U. School of Computer Science and Engineering Tamil Nadu Chennai India

ISBN: (纸本)9798350370249

Abstmet -Reliable sensor measurements are essental for autonomous navigation, control systems and various other applications. We studied a model of a gesture controlled wheelchair. A key finding is that the processing unit must be able to filter out noise so that the efficacy of the wearable device is improved. In this paper, we demonstrate the application of the Kalman filter to combine measurements in the presence of noise or drift, to calculate the orientation from the inertial measurement unit, MPU6050. We report the behoviour of the gyroscope and accelerometer by measuring their output values individually first, and then passing them through the Kalman Filter. Alternatives such as the recursive average and moving average filters are also discussed. The characteristics of the filter for different parameters such as the standard deviation are plotted. Finally, the Kalman gain convergence for a steady state is verified. © 2024 IEEE.

关键词： Adaptive filters Adaptive signal processing Gesture recognition IIR filters Kalman filters Kynwrds - Control systems signal processing algorithms

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：