Article
Computer Science, Artificial Intelligence
Jinrae Kim, Youdan Kim
Summary: This article introduces parameterized max-affine and parameterized log-sum-exp networks for general decision-making problems. These approximators generalize existing convex approximators by considering function arguments and replacing network parameters with continuous functions. The universal approximation theorem is proven, showing that these networks are shape-preserving universal approximators. Practical guidelines for incorporating deep neural networks are provided and numerical simulations demonstrate the performance of the proposed approximators, with PLSE outperforming other approximators in terms of minimizer and optimal value errors in high-dimensional cases.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Automation & Control Systems
Xinlei Yi, Xiuxian Li, Tao Yang, Lihua Xie, Tianyou Chai, Karl Henrik Johansson
Summary: This paper explores distributed bandit online convex optimization with time-varying coupled inequality constraints, focusing on the repeated game between learners and an adversary. By optimizing the global loss functions and coupled constraint functions in multiple iterations, the algorithms are able to achieve sublinear expected regret and constraint violation.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2021)
Article
Engineering, Aerospace
Peng Lu, Xinfu Liu, Runqiu Yang, Zhiguo Zhang
Summary: In this article, a method for optimizing ascent trajectory in the atmospheric phase is proposed. The problem is decomposed into three subproblems and convexification techniques are used to handle each subproblem. The proposed approach retains most of the nonlinearity of the original problem and can converge quickly even with rough initial state and control profiles.
IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Tianyi Liu, Andreas M. Tillmann, Yang Yang, Yonina C. Eldar, Marius Pesavento
Summary: This paper investigates the phase retrieval problem with dictionary learning and proposes two complementary algorithms for blind channel estimation, which are evaluated in a multi-antenna random access network.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Article
Mathematics, Applied
Jourdain Lamperski, Oleg A. Prokopyev, Luca G. Wrabetz
Summary: This paper studies the min-max-min optimization problem with smooth and strongly convex objectives. By connecting the problem to the k-center problem and min-max-min robust optimization, a generalized approximation algorithm is proposed and compared with an exact method.
SIAM JOURNAL ON OPTIMIZATION
(2023)
Article
Engineering, Electrical & Electronic
Chen Zhang Feng, Wen Tao Li, Can Cui, Yong Qiang Hei, Jin Chao Mou, Xiao Wei Shi
Summary: This letter proposes a novel static-dynamic convex optimization algorithm for array synthesis. The algorithm is divided into static and dynamic successive optimization stages and uses a synthetic direction graph and shrinkage factor and crossover operator to ensure the continuity of the solution process and approach feasible solutions.
IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS
(2022)
Article
Operations Research & Management Science
Hsu Kao, Vijay Subramanian
Summary: Distributed optimization has many applications and convex optimization methods are well studied. However, non-convex optimization methods are not well understood. This paper proposes a new non-convex distributed optimization framework and extends it to resource allocation problems, reducing communication and memory complexity while relaxing the gradient assumption.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2023)
Article
Computer Science, Software Engineering
Daniel Doerfler, Andreas Loehne, Christopher Schneider, Benjamin Weissing
Summary: The algorithm presented aims to approximately solve bounded convex vector optimization problems by providing both outer and inner polyhedral approximations, improving approximation accuracy. Compared to previous algorithms, a new selection rule is proposed for potentially achieving the same approximation quality faster.
OPTIMIZATION METHODS & SOFTWARE
(2022)
Article
Mathematics, Applied
Vidhi Zala, Robert M. Kirby, Akil Narayan
Summary: Finite element simulations are widely used to solve partial differential equations modeling various physical, chemical, and biological phenomena. However, the discretized solutions often do not satisfy necessary physical properties such as positivity or monotonicity. Researchers propose embedding these structural solution properties as additional constraints in a convex optimization process to compute solutions that satisfy convex constraints.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2021)
Article
Management
Ying-Chao Hung, Horace PakHai Lok, George Michailidis
Summary: In this paper, a general electric vehicle charging system with stochastic demand, demand request locations, and predetermined charging facilities is considered. The objective is to design a routing strategy that accommodates demand-request dynamics, satisfies stability constraints, and minimizes the mean response time. A flexible and measurement-based routing policy called partition-based random routing (PBRR) is introduced, and the performance measure is formulated as a constrained optimization problem. A surrogate optimization algorithm is proposed for finding the optimal routing solution. Numerical results demonstrate the satisfactory performance of the developed routing strategy and its fast implementation.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2022)
Review
Automation & Control Systems
Alan Yang, Stephen Boyd
Summary: This article proposes a method for designing policies for convex stochastic control problems. The method evaluates control policies by using quadratic approximate value functions as a substitute for the true value function. The method can yield a good approximate value function with few samples and little hyperparameter tuning.
ANNUAL REVIEWS IN CONTROL
(2023)
Article
Computer Science, Software Engineering
Dmitry Kamzolov, Pavel Dvurechensky, Alexander Gasnikov
Summary: In this paper, we propose a new first-order method for minimizing a convex function, which combines the characteristics of Universal Gradient Method and Intermediate Gradient Method. By automatically adjusting itself to the local level of smoothness, it achieves better performance. The restart technique can be used to accelerate convergence speed under the additional assumption of strong convexity.
OPTIMIZATION METHODS & SOFTWARE
(2021)
Article
Computer Science, Interdisciplinary Applications
Ming Zhou, Ole Sigmund
Summary: The paper discusses Sigmund's 2001 educational paper with a self-contained 99-line MATLAB code, which has had a far-reaching impact on teaching and research of topology optimization. The goal of the paper is to provide clarity to the theoretical foundation and enable students to learn the complete iterative optimization solution with minimum additional effort.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2021)
Article
Automation & Control Systems
Runyue Fang, Dequan Li, Xiuyu Shen
Summary: This paper proposes a distributed online adaptive subgradient learning algorithm called DAdaxBound, which can effectively handle optimization problems in time-varying networks and has good performance on convex and potentially nonsmooth objective functions.
IET CONTROL THEORY AND APPLICATIONS
(2022)
Article
Automation & Control Systems
Yiyue Chen, Abolfazl Hashemi, Haris Vikalo
Summary: In this article, a communication-efficient decentralized optimization scheme is proposed for time-varying directed networks using sparsification, gradient tracking, and variance reduction. The scheme achieves an accelerated linear convergence rate for smooth and strongly convex objective functions, making it the first of its kind for time-varying directed networks. Experimental results on synthetic and real datasets demonstrate the efficacy of the proposed scheme.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2022)
Article
Engineering, Electrical & Electronic
Xinghua Liu, Jianwei Guan, Rui Jiang, Xiang Gao, Badong Chen, Shuzhi Sam Ge
Summary: This paper introduces a state estimation approach for nonlinear systems with unknown inputs, utilizing statistical linearization and weighted least squares to linearize the nonlinear state and measurement equations, as well as estimate unknown inputs. The method also incorporates multiple suboptimal fading factors to improve tracking ability for inaccuracies in the system model and abrupt changes in state variables caused by unknown inputs. Furthermore, the approach employs unbiased minimum variance estimation and singular value decomposition to enhance algorithm stability, with simulated results confirming its effectiveness.
IET SIGNAL PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Ben Yang, Xuetao Zhang, Badong Chen, Feiping Nie, Zhiping Lin, Zhixiong Nan
Summary: Despite the widespread attention to multi-view clustering for its superior performance, it still faces challenges such as high computational cost and complex noises. This paper introduces ECMC algorithm to enhance efficiency and robustness by utilizing correntropy and NMF, showing faster speed and better performance than other state-of-the-art algorithms.
Article
Computer Science, Artificial Intelligence
Yunfei Zheng, Badong Chen, Shiyuan Wang, Weiqun Wang, Wei Qin
Summary: This article introduces the kernel-based extreme learning machine (KELM) and its outstanding performance in addressing regression and classification problems. To improve the robustness of KELM, a mixture correntropy-based KELM (MC-KELM) is proposed, which adopts a maximum mixture correntropy criterion as the optimization criterion. Additionally, an online sequential version of MC-KELM (MCOS-KELM) is developed to handle sequentially arriving data. Experimental results demonstrate the superior performance of the new methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Ben Yang, Xuetao Zhang, Feiping Nie, Badong Chen, Fei Wang, Zhixiong Nan, Nanning Zheng
Summary: This paper proposes an orthogonal conceptual factorization (OCF) model and an efficient correntropy-based clustering algorithm (ECCA) to cluster large amounts of unlabeled data. The OCF model restricts the degree of freedom of matrix factorization, while ECCA performs OCF on an anchor graph to improve clustering efficiency and robustness.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Binghui Li, Badong Chen
Summary: This paper introduces an adaptive RRT-Connect (ARRT-Connect) planning method, which handles narrow passage environments and retains the path planning ability of RRT algorithms in other environments.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2022)
Article
Automation & Control Systems
Lujuan Dang, Badong Chen, Yulong Huang, Yonggang Zhang, Haiquan Zhao
Summary: The paper proposes a robust nonlinear Kalman filter called MEEF-CKF, which exhibits strong robustness against complex non-Gaussian noises. It addresses the issue of biased estimates in the traditional CKF when the INS/GPS system faces non-Gaussian disturbances. The MEEF-CKF operates major steps including regression model construction, robust state estimation, and free parameters optimization for enhanced robustness.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2022)
Article
Computer Science, Artificial Intelligence
Weiwei Shi, Yihong Gong, Badong Chen, Xinhong Hei
Summary: The proposed transductive semisupervised deep hashing (TSSDH) method effectively trains deep convolutional neural network (DCNN) models with both labeled and unlabeled training samples, outperforming representative semisupervised deep hashing methods in terms of image retrieval accuracies under the same number of labeled training samples.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Hao Wu, Nanning Zheng, Badong Chen
Summary: Decoding the content in neural activity is crucial for investigating cognitive functions of the human brain. Traditional voxelwise encoding methods ignore the interactions between voxels and are sensitive to noise. This study proposes a feature-specific denoise method to improve decoding performance by reducing the feature-irrelevant component in voxels.
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Ben Yang, Xuetao Zhang, Zhiping Lin, Feiping Nie, Badong Chen, Fei Wang
Summary: This paper proposes an efficient and robust multi-view clustering algorithm with anchor graph regularization (ERMC-AGR). By designing a novel anchor graph regularization and using nonnegative matrix factorization based on correntropy criterion, the algorithm improves clustering results, reduces computational complexity, and shows promising performance in handling complex noises and outliers.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Yuanhao Li, Badong Chen, Natsue Yoshimura, Yasuharu Koike
Summary: The instantiation of MEE on robust classification is lacking in the literature, thus a specific criterion called Restricted MEE (RMEE) is proposed with optimization and convergence analysis. Experimental results demonstrate the superior robustness of RMEE on synthetic and real datasets, as well as its practical impact on noisy electroencephalogram dataset.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Ziru Wang, Jiawen Liu, Yongqiang Ma, Badong Chen, Nanning Zheng, Pengju Ren
Summary: Perturbation contributes to the stability and exploration ability of neural systems through parameter changes and behavior modifications. In artificial neural systems, perturbation can make network structures more robust and diverse. For spiking neural networks (SNN), temporal perturbation may play an active role in performance improvement.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Automation & Control Systems
Mingming Bai, Yulong Huang, Badong Chen, Yonggang Zhang
Summary: A novel normal-skew mixture (NSM) distribution is introduced in this article to model various noise distributions. Utilizing this distribution, a new robust Kalman filtering framework is developed and several exemplary robust Kalman filters are derived based on it. The proposed framework is proven to have superior performance in target tracking simulation.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Automation & Control Systems
Badong Chen, Yuqing Xie, Xin Wang, Zejian Yuan, Pengju Ren, Jing Qin
Summary: Correntropy is a novel similarity measure used in machine learning and signal processing. The concept of multikernel correntropy (MKC) has been proposed to improve learning performance. Experimental results demonstrate that MKC outperforms traditional methods.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Acoustics
Yingying Zhu, Haiquan Zhao, Xiaoqiong He, Zeliang Shu, Badong Chen
Summary: This paper proposes a new algorithm to deal with impulsive noise and reduce computational complexity. By designing a cascaded random Fourier filter model, the cascaded RF-FxGHSF algorithm is derived and the steady-state convergence conditions are analyzed. Experimental results verify the convergence ability of the algorithm.
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2022)
Article
Engineering, Electrical & Electronic
Lujuan Dang, Wanli Wang, Badong Chen
Summary: Dynamic state estimation of a power system is essential and nonlinear Kalman filters have been identified as versatile tools for this task. However, the presence of non-Gaussian noise makes the available observations inaccurate, degrading the precision and robustness of the filtering process. To address this issue, a novel method based on a square root unscented Kalman filter is proposed, which can directly act on measurement to weight error covariance and noise variance. Simulation results show that the proposed method achieves significantly improved filtering performance compared to other related nonlinear Kalman filters.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)