检索结果-内蒙古大学图书馆

Stagewise Training With Exponentially Growing Training Sets

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2025年第4期36卷 6148-6158页

作者： Gu, Bin Alquabeh, Hilal de Vazelhes, William Huo, Zhouyuan Huang, Heng Jilin Univ Sch Artificial Intelligence Changchun 130012 Peoples R China Mohamed bin Zayed Univ Artificial Intelligence Dept Machine Learning Abu Dhabi U Arab Emirates Univ Pittsburgh Dept Elect & Comp Engn Pittsburgh PA 15260 USA Univ Maryland Dept Comp Sci College Pk MD 20742 USA

In the world of big data, training large-scale machine learning problems has gained considerable attention. Numerous innovative optimization strategies have been presented in recent years to accelerate the large-scale training process. However, the possibility of further accelerating the training process of various optimization algorithms remains an unresolved subject. To begin addressing this difficult problem, we exploit the researched findings that when training data are independent and identically distributed, the learning problem on a smaller dataset is not significantly different from the original one. Upon that, we propose a stagewise training technique that grows the size of the training set exponentially while solving nonsmooth subproblem. We demonstrate that our stagewise training via exponentially growing the size of the training sets (STEGSs) are compatible with a large number of proximal gradient descent and gradient hard thresholding (GHT) techniques. Interestingly, we demonstrate that STEGS can greatly reduce overall complexity while maintaining statistical accuracy or even surpassing the intrinsic error introduced by GHT approaches. In addition, we analyze the effect of the training data growth rate on the overall complexity. The practical results of applying l(2,1) -and l(0) -norms to a variety of large-scale real-world datasets not only corroborate our theories but also demonstrate the benefits of our STEGS framework.

关键词： Training Optimization Computational complexity Upper bound approximation algorithms Training data Risk minimization Gradient hard thresholding (GHT) algorithms proximal algorithms stagewise training strategy stochastic optimization

来源：评论

学校读者我要写书评

暂无评论

A Terminal Constraint Set-Dependent Mixed Time/Event-Triggered Approach to Multistep Fuzzy MPC

引用

IEEE TRANSACTIONS ON FUZZY SYSTEMS 2025年第2期33卷 606-620页

作者： Wei, Fan Wan, Xiongbo Zhang, Chuan-Ke Wang, Leimin China Univ Geosci Sch Automat Wuhan 430074 Peoples R China Hubei Key Lab Adv Control & Intelligent Automat Co Wuhan 430074 Peoples R China Minist Educ Engn Res Ctr Intelligent Technol Geoexplorat Wuhan 430074 Peoples R China

In this article, the problem of multistep model predictive control (MPC) under a mixed time/event-triggered mechanism (MTEM) is investigated for fuzzy systems with hard constraints. To improve its flexibility in adjusting releases, this MTEM incorporates the information from the fuzzy membership functions and the terminal constraint set (TCS). In the online control unit, the time-triggered way is first implemented to steer the system state into the TCS quickly, and then in the offline control unit, the involved event-triggered pattern plays its role in conserving resources while ensuring the control performance. Two offline optimization problems (OPs) are presented to design the TCS and the approximate robust one step sets, respectively, and an online OP is given to design the control laws to steer the system state into TCS. Based on these three OPs, we propose an MTEM-based multistep fuzzy MPC algorithm and demonstrate the feasibility of the algorithm together with the input-to-state stability of the closed-loop system. Three examples are given to verify the effectiveness of the proposed method and its superiority in saving communication and computing resources while ensuring control performance.

关键词： Fuzzy systems Event detection Vectors Symmetric matrices Predictive models Predictive control Optimization Prediction algorithms Nonlinear systems approximation algorithms Mixed time/event-triggered mechanism (MTEM) multistep model predictive control (MPC) T-S fuzzy systems terminal constraint set (TCS)

来源：评论

学校读者我要写书评

暂无评论

Adaptive Point Mass Filter and Its Application in Terrain Matching Navigation

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2025年 74卷

作者： Zhao, Shengwu Deng, Zhihong Zhang, Wenzhe Wang, Yu Beijing Inst Technol Sch Automat Beijing 100081 Peoples R China Beijing Inst Technol Natl Key Lab Intelligent Control & Decis Complex S Beijing 100081 Peoples R China

Terrain matching algorithm is the core part of terrain-aided inertial navigation system (INS), and point mass filter (PMF) is often used as a terrain matching algorithm. In PMF, the construction method and calculation amount of the point mass set are important parts which affect the accuracy. To comprehensively consider the construction of point mass set and calculation amount, this article proposes an adaptive construction method of point mass set based on probability distribution. According to the characteristics of the probability distribution to be approximated, the method transforms the preset point mass set to obtain point mass set of the current moment. Aiming at the problem of calculation error, which is caused by the process noise intensity being smaller than the resolution of the point mass set, a local convolution method is proposed in this article, in which the convolution result is obtained by subdividing the local area near the point mass set. Finally, the framework of the proposed adaptive PMF (APMF) in terrain matching is given. Numerical simulations show the superiority of the proposed APMF, and the experiment shows the effectiveness of the proposed method in terrain-aided INS.

关键词： Convolution approximation algorithms Noise Probability distribution Navigation Underwater vehicles Transient analysis Training Posterior probability Position measurement Adaptive point mass set local convolution point mass filter (PMF) terrain matching navigation

来源：评论

学校读者我要写书评

暂无评论

Efficient Change-Point Detection Over Fully Decentralized Wireless Networks With Low Communication Rate

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2025年第1期74卷 1626-1642页

作者： Jiao, Yuchen Meng, Linghang Li, Ying Yu, Quan Gu, Yuantao Beijing Inst Technol Sch Informat & Elect Beijing 100081 Peoples R China China Telecom Cloud Comp Res Inst Beijing 100033 Peoples R China Peng Cheng Lab Shenzhen 518066 Peoples R China Tsinghua Univ Beijing Natl Res Ctr Informat Sci & Technol Dept Elect Engn Beijing 100084 Peoples R China

Change-point detection (CPD), also known as quickest detection, aims to detect an abrupt change in the environment according to its observations as quickly as possible. Its applications in networks, such as wireless sensor networks and vehicular networks, motivate the study of distributed CPD, also known as multi-sensor CPD. In this context, multiple sensors monitor the environment via observations, but only an unknown subset of sensors is affected by the abrupt change. In this paper, we propose a novel distributed CPD algorithm, which simultaneously enjoys the advantages of being free of a fusion center, having low energy cost, being robust to the number of affected sensors, and suitable for a general data model. In particular, the proposed algorithm incorporates the thresholding technique into the a decentralized variant of cumulative sum (CUSUM) procedure. Sensors exchange information only when a designed local statistic exceeds a threshold, thereby reducing the number of communications and the energy cost. In addition, the proposed algorithm approximates the sum-CUSUM procedure and achieves robustness to the number of affected sensors. Theoretical analysis guarantees that the proposed algorithm achieves an exponential relationship between two detection performance criteria, the average running length (ARL), which measures the false alarm rate, and the expected detection delay (EDD). This result holds regardless of the number of the affected sensors. Furthermore, the thresholding operation only has a mild influence on the detection performance. Finally, we conduct simulations to verify the efficiency and effectiveness of the proposed algorithm. We also apply it to distributed source detection task in multiple-antenna wireless networks as a potential application.

关键词： Sensors Costs Sensor fusion Robustness approximation algorithms Resilience Quantization (signal) Change-point detection (CPD) communication cost CUSUM fully decentralized

来源：评论

学校读者我要写书评

暂无评论

Average Controllability of Complex Networks With Exponential Degree Distribution

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS 2025年第4期72卷 1813-1821页

作者： Xiang, Linying Zhu, Zeya Zhu, Jiawei Yao, Shuwei Chen, Fei Tiangong Univ Sch Artificial Intelligence Tianjin 300387 Peoples R China Tiangong Univ Tianjin Key Lab Intelligent Control Elect Equipmen Tianjin 300387 Peoples R China Shanghai Jiao Tong Univ Sch Elect Informat & Elect Engn Shanghai 200240 Peoples R China Nankai Univ Coll Artificial Intelligence Tianjin 300350 Peoples R China

This paper investigates the average controllability of Laplacian dynamical networks characterized by an exponential degree distribution. We introduce a novel configuration network model with an exponential degree distribution, incorporating a degree distribution parameter to adjust the heterogeneity of the distribution. We thoroughly examine the impact of degree distribution and degree correlation on average controllability. Our results reveal that increased heterogeneity in degree distribution tends to enhance average controllability, particularly when edges connect nodes with higher degrees. Moreover, networks with high assortativity exhibit improved average controllability. Building on these findings, we propose two effective strategies for optimizing average controllability. These strategies offer valuable guidance for the design of optimal complex networks in practical settings.

关键词： Controllability Laplace equations Eigenvalues and eigenfunctions Complex networks Correlation Vectors Dynamical systems approximation algorithms Power system dynamics Circuits Complex network average controllability Laplacian dynamics exponential degree distribution degree correlation

来源：评论

学校读者我要写书评

暂无评论

Curriculum Learning algorithms for Reward Weighting in Sparse Reward Robotic Manipulation Tasks

引用

IEEE ACCESS 2025年 13卷 45544-45558页

作者： Fele, Benjamin Babic, Jan Jozef Stefan Inst Ljubljana 1000 Slovenia

Robotic learning from sparse rewards can be a considerable challenge due to large amounts of data required for mastering a task. We explore the application of curriculum learning (CL) algorithms for automatic reward weighting to tackle learning from sparse rewards in robotic pick-and-place and stacking tasks. We take several state-of-the-art CL algorithms that were originally designed to generate curriculum by manipulating the environment and appropriate them to weigh multiple sparse reward functions instead. The reward functions are chosen in a way that facilitates staged learning of the task, and the two robotic tasks are designed so that the agent learns to generalize to any initial and goal object position in the scene. The results of our three implemented CL algorithms show large improvement over the naive and state-of-the-art baselines in terms of speed of convergence to a successful policy in experiments with multiple task variations. Various generalization tests showcase some strengths and weaknesses of our approach. Inspection of changes in reward weight values during training further reveals varying curricula generated by the employed approaches, and showcases shifting emphasis from auxiliary to the main reward as the training progresses.

关键词： Robots Reinforcement learning Training Stacking Heuristic algorithms Convergence approximation algorithms Zinc Space exploration Prediction algorithms Curriculum learning robotic manipulation reinforcement learning reward shaping robotic learning sparse reward

来源：评论

学校读者我要写书评

暂无评论

Generalized Policy Improvement algorithms With Theoretically Supported Sample Reuse

引用

IEEE TRANSACTIONS ON AUTOMATIC CONTROL 2025年第2期70卷 1236-1243页

作者： Queeney, James Paschalidis, Ioannis Ch. Cassandras, Christos G. Mitsubishi Elect Res Labs Cambridge MA 02139 USA Boston Univ Dept Elect & Comp Engn Boston MA 02215 USA Boston Univ Div Syst Engn Boston MA 02215 USA

We develop a new class of model-free deep reinforcement learning algorithms for data-driven, learning-based control. Our Generalized Policy Improvement algorithms combine the policy improvement guarantees of on-policy methods with the efficiency of sample reuse, addressing a tradeoff between two important deployment requirements for real-world control: 1) practical performance guarantees;and 2) data efficiency. We demonstrate the benefits of this new class of algorithms through extensive experimental analysis on a broad range of simulated control tasks.

关键词： approximation algorithms Optimization TV Training Task analysis Trajectory Heuristic algorithms Policy improvement policy optimization reinforcement learning (RL) sample reuse

来源：评论

学校读者我要写书评

暂无评论

Achieving High-Quality Image Stylization Using Hierarchical Sensibility Boost, Visual Style Attention, and Covariance Evolution Blend

引用

IEEE ACCESS 2025年 13卷 5937-5948页

作者： Gao, Ya Shandong Management Univ Sch Art Jinan 250357 Peoples R China

In recent years, image stylization has gained significant attention due to its ability to transform visual styles from reference images to content images, transforming ordinary photographs into sensuous creations. Traditional methods, however, struggle to retain fine-grained details and often exhibit issues such as discordant textures and noticeable artifacts. This paper proposes a novel approach for high-quality image stylization by introducing a method that integrates Hierarchical Sensibility Boost (HSB), Visual Style Attention (VSA), and Covariance Evolution Blend (CovEB) blocks. The proposed method addresses the challenges of preserving content construction while enhancing stylistic representations to achieve visually pleasing and realistic results. Specifically, the HSB block captures and amplifies style representations at different sizes, the VSA block uses channel attention mechanisms to dynamically boost style representations, and the CovEB block aligns second-order statistical data between style and content representations to ensure coherent stylization. Comprehensive experiments conducted on MS-COCO and WikiArt datasets demonstrate that the proposed method produces stylized images with superior visual sensibility, minimal artifacts, and improved efficiency compared to existing techniques. Quantitative evaluations confirm that our method excels in style similarity, global effects, and local pattern retention, while maintaining competitive performance in content fidelity and runtime efficiency.

关键词： Visualization Feature extraction Transforms Image reconstruction Training Image color analysis Decoding Painting Data mining approximation algorithms Neural network stylization visual sensibility attention

来源：评论

学校读者我要写书评

暂无评论

Sparse Aperture ISAR Autofocusing and Imaging Algorithm Based on Log-Sum Regularization

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2025年 63卷

作者： Zhou, Xin Zhao, Zhiyuan Shen, Jianjun Liu, Zhen Liu, Yongxiang Xue, Bin Natl Univ Def Technol Coll Informat & Commun Wuhan 430010 Peoples R China Natl Univ Def Technol Coll Elect Sci & Technol Changsha 410037 Peoples R China

Mathematically, the autofocusing and imaging model for sparse aperture inverse synthetic aperture radar (ISAR) has an infinite number of solutions, even with the addition of some sparsity constraints. What relationship exists between these infinite solutions should be answered. In addition, some existing approaches suffer from troublesome manual fine-tuning parameters, or high algorithmic complexity, or low reconstruction accuracy. To deal with the above problems, a sparse aperture ISAR autofocusing and imaging method combining phase estimation and log-sum minimization is proposed in this article, which has high estimation accuracy and computational efficiency, and avoids complicated manual parameter adjustment. Focusing on the model with a phase diagonal matrix and log-sum function, we reveal why there are countless solutions and what relationships exist between them. These features are shared by other models including different regularization functions. Without sparsity prior, the regularization parameter of the algorithm is changed automatically at each iteration according to the ratio of the selected element to the maximum absolute value element in an auxiliary matrix, which controls that only a few elements participate in the computation at each iteration. Coupled with the situation that computationally heavy matrix inversion operations are not required, the calculation speed of the approach is greatly boosted. To show the stability of the algorithm, its convergence is proven without any assumptions. Experiments based on both simulation and measurement data demonstrate that compared with the recently proposed algorithms by others, the proposed algorithm can achieve well-focused ISAR images in less than one second and is very efficient to implement.

关键词： Imaging Radar imaging Apertures Image reconstruction Convergence approximation algorithms Sparse matrices Phase estimation Minimization Computational modeling Compressive sensing (CS) inverse synthetic aperture radar (ISAR) image log-sum minimization phase estimation sparse aperture

来源：评论

学校读者我要写书评

暂无评论

Online Actuator Selection and Controller Design for Linear Quadratic Regulation With an Unknown System Model

引用

IEEE TRANSACTIONS ON AUTOMATIC CONTROL 2025年第1期70卷 18-33页

作者： Ye, Lintao Chi, Ming Liu, Zhi-Wei Gupta, Vijay Huazhong Univ Sci & Technol Key Laboratoryof Image Proc & Intelligent Control Minist Educ Wuhan 430074 Peoples R China Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Wuhan 430074 Peoples R China Purdue Univ Elmore Family Sch Elect & Comp Engn W Lafayette IN 47907 USA

In this article, we study the simultaneous actuator selection and controller design problem for linear quadratic regulation with Gaussian noise over a finite horizon of length $T$ and unknown system model. We consider episodic and nonepisodic settings of the problem and propose online algorithms that specify both the sets of actuators to be utilized under a cardinality constraint and the controls corresponding to the sets of selected actuators. In the episodic setting, the interaction with the system breaks into $N$ episodes, each of which restarts from a given initial condition and has length $T$. In the nonepisodic setting, the interaction goes on continuously. Our online algorithms leverage a multiarmed bandit algorithm to select the sets of actuators and a certainty equivalence approach to design the corresponding controls. We show that our online algorithms yield $\sqrt{N}$-regret for the episodic setting and $T<^>{2/3}$-regret for the nonepisodic setting. We extend our algorithm design and analysis to show scalability with respect to both the total number of candidate actuators and the cardinality constraint. We numerically validate our theoretical results.

关键词： Actuators Trajectory Regulation approximation algorithms Perturbation methods Costs Additives Large-scale system multiarmed bandit problem optimal control reinforcement learning stochastic system

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：