检索结果-内蒙古大学图书馆

An Artificial Neural Network Based Strategy for Commutation Failure Forecasting in LCC-HVDC Transmission Networks

学校读者我要写书评

暂无评论

An Artificial Neural Network Based Strategy for Commutation ...

2023 IEEE International Conference on Energy Technologies for Future Grids, ETFG 2023

作者： Mirsaeidi, Sohrab Shang, Fuxiang Ghaffari, Kimia He, Jinghan Said, Dalila Mat Muttaqi, Kashem M. Beijing Jiaotong University School of Electrical Engineering Beijing China Beijing Jiaotong University School of Economics and Management Beijing China School of Electrical Engineering Johor Malaysia University of Wollongong School of Electrical Computer and Telecommunications Engineering Wollongong Australia

ISBN: (纸本)9781665471640

Commutation failure (CF) is a major challenge associated with line-commutated converter high voltage dc (LCC-HVdc) systems. Faults at the inverter ac systems are regarded as main cause of commutation failure. The present study proposes a method for forecasting the commutation failure based on a feed-forward neural network (FFNN) combined with optimization algorithms and principal component analysis (PCA). To fulfill this, first, a two-terminal HVdc system is designed, and then, a large number of three-phase faults with different fault inductances are applied on the inverter ac side. In each fault case, 100 samples of the dc current signal in the interval between fault and commutation failure initial times are taken with the sampling frequency of 1 MHz. Subsequently, with the aim of dimensionality reduction, the samples are preprocessed by the PCA before serving as inputs of the neural network. Finally, various metaheuristic optimization methods, i.e. particle swarm optimization (PSO), grey wolf optimizer (GWO), simulated annealing (SA), and genetic algorithm (GA) are utilized to determine the most appropriate structure for the neural network in terms of the number of neurons in each hidden layer, and the type of transfer function (TF) in hidden layers and output layer. The simulation results indicate that the optimized artificial neural network can predict the commutation failure occurrence with an accuracy of roughly 90%. © 2023 IEEE.

关键词： artificial neural network Commutation failure forecasting optimization algorithms principal component analysis

Extensive Benchmarking of Metaheuristics Using Real-World Constrained Multi-Objective optimization Problems

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Mukhopadhyay, Somnath Sarkar, Sunita Das, Devangshu Phukan, Madhurjya Purkyastha, Saptarshi Dept. of Computer Science & Engineering Assam University Silchar India

The current optimization and Metaheuristics literature is booming with contributions. While this surge fosters advancements in the field, it has also made it into an over-saturated one. Consequently, people are often faced with a plethora of optimization algorithms to choose from. An extensive analysis of the algorithms is essential for better understanding of the abilities of these algorithms. But most of the studies conducted only employ Synthetic Benchmark Problems (SBPs), which can cause overestimation of the algorithms because of their simplified structure, the algorithms also struggle when they are used in real-world contexts. To fill the gap between the theory and the practical, we test several Metaheuristics algorithm using Real World Constrained Multi-Objective optimization Problems (RWCMOPs) and rank them using various performance measures like Hypervolume Indicator (HV), Feasibilty Rate (FR) and Total Wall-Clock Time. An analysis using Real World problems and multiple performance indicators will help in better understanding of the capabilities of these Metaheuristic algorithms. © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Effective Edge Centrality via Neighborhood-based optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yang, Renchi Department of Computer Science Hong Kong Baptist University Hong Kong

Given a network G, edge centrality is a metric used to evaluate the importance of edges in G, which is a key concept in analyzing networks and finds vast applications involving edge ranking. In spite of a wealth of research on devising edge centrality measures, they incur either prohibitively high computation costs or varied deficiencies that lead to sub-optimal ranking quality. To overcome their limitations, this paper proposes ECHO, a new centrality measure for edge ranking that is formulated based on neighborhood-based optimization objectives. We provide in-depth theoretical analyses to unveil the mathematical definitions and intuitive interpretations of the proposed ECHO measure from diverse aspects. Based thereon, we present three linear-complexity algorithms for ECHO estimation with nontrivial theoretical accuracy guarantees for centrality values. Extensive experiments comparing ECHO against six existing edge centrality metrics in graph analytics tasks on real networks showcase that ECHO offers superior practical effectiveness while offering high computation efficiency. Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Adaptive Backtracking For Faster optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Cavalcanti, Joao V. Lessard, Laurent Wilson, Ashia C. MIT United States Northeastern University United States

Backtracking line search is foundational in numerical optimization. The basic idea is to adjust the step size of an algorithm by a constant factor until some chosen criterion (e.g. Armijo, Goldstein, Descent Lemma) is satisfied. We propose a new way for adjusting step sizes, replacing the constant factor used in regular backtracking with one that takes into account the degree to which the chosen criterion is violated, without additional computational burden. For convex problems, we prove adaptive backtracking requires fewer adjustments to produce a feasible step size than regular backtracking does for two popular line search criteria: the Armijo condition and the descent lemma. For nonconvex smooth problems, we additionally prove adaptive backtracking enjoys the same guarantees of regular backtracking. Finally, we perform a variety of experiments on over fifteen real world datasets, all of which confirm that adaptive backtracking often leads to significantly faster optimization. © 2024, CC BY.

关键词： optimization algorithms

Data-Driven Discrete Global optimization Using an Information Mining Strategy for Costly Black-Box Problems

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Wang, Shengfa Dong, Huachao Wang, Peng Zhu, Haijia Wen, Zhiwen Xiang, Junyu School of Marine Science and Technology Northwestern Polytechnical University Xi’an710068 China University of Victoria Canada Xi’an Precision Machinery Research Institute Xi’an710077 China

Surrogate-assisted optimization algorithms are frequently applied to handle costly black-box optimization problems. However, most of the research is on continuous problems and there are fewer studies on solving discrete problems. In addition, there is underutilization of information from surrogate models and discrete data when solving discrete problems. To address these issues, this paper proposes a data-driven discrete global optimization algorithm (DDGO). The algorithm uses a two-stage optimization framework to incorporate efficient continuous optimization modules and fully exploit the information of the discrete problem itself and the surrogate model. In the first stage, a multiple starting points selection strategy is used to obtain valuable starting points for multi-start optimization. In the second stage, a discrete neighborhoods selection strategy achieves the selection from candidate continuous solutions to optimal discrete solutions. The DDGO algorithm is tested on 20 representative benchmark cases and can find better solutions compared to other 7 algorithms. Additionally, the algorithm is employed in the structural optimization design of a blended-wing-body underwater glider, which proves that the DDGO algorithm has a good capability for the solution of discrete black-box problems. © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Variable Metric Evolution Strategies for High-dimensional Multi-Objective optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Glasmachers, Tobias Institute for Neural Computation Ruhr-University Bochum Germany

We design a class of variable metric evolution strategies well suited for high-dimensional problems. We target problems with many variables, not (necessarily) with many objectives. The construction combines two independent developments: efficient algorithms for scaling covariance matrix adaptation to high dimensions, and evolution strategies for multi-objective optimization. In order to design a specific instance of the class we first develop a (1+1) version of the limited memory matrix adaptation evolution strategy and then use an established standard construction to turn a population thereof into a state-of-the-art multi-objective optimizer with indicator-based selection. The method compares favorably to adaptation of the full covariance matrix. Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

OPTIMAL STRONG REGRET AND VIOLATION IN CONSTRAINED MDPS VIA POLICY optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Stradi, Francesco Emanuele Castiglioni, Matteo Marchesi, Alberto Gatti, Nicola Politecnico di Milano Italy

We study online learning in constrained MDPs (CMDPs), focusing on the goal of attaining sublinear strong regret and strong cumulative constraint violation. Differently from their standard (weak) counterparts, these metrics do not allow negative terms to compensate positive ones, raising considerable additional challenges. Efroni et al. (2020) were the first to propose an algorithm with sublinear strong regret and strong violation, by exploiting linear programming. Thus, their algorithm is highly inefficient, leaving as an open problem achieving sublinear bounds by means of policy optimization methods, which are much more efficient in practice. Very recently, Müller et al. (2024) have partially addressed this problem by proposing a policy optimization method that allows to attain Oe(T 0.93) strong regret/violation. This still leaves open the question of whether optimal bounds are achievable by using an approach of this kind. We answer such a question affirmatively, by providing an efficient policy optimization algorithm with Oe(√T) strong regret/violation. Our algorithm implements a primal-dual scheme that employs a state-of-the-art policy optimization approach for adversarial (unconstrained) MDPs as primal algorithm, and a UCB-like update for dual variables. Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Geometric-Averaged Preference optimization for Soft Preference Labels

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Furuta, Hiroki Lee, Kuang-Huei Gu, Shixiang Shane Matsuo, Yutaka Faust, Aleksandra Zen, Heiga Gur, Izzeddin Google DeepMind United Kingdom The University of Tokyo Japan

Many algorithms for aligning LLMs with human preferences assume that human preferences are binary and deterministic. However, human preferences can vary across individuals, and therefore should be represented distributionally. In this work, we introduce the distributional soft preference labels and improve Direct Preference optimization (DPO) with a weighted geometric average of the LLM output likelihood in the loss function. This approach adjusts the scale of learning loss based on the soft labels such that the loss would approach zero when the responses are closer to equally preferred. This simple modification can be easily applied to any DPO-based methods and mitigate over-optimization and objective mismatch, which prior works suffer from. Our experiments simulate the soft preference labels with AI feedback from LLMs and demonstrate that geometric averaging consistently improves performance on standard benchmarks for alignment research. In particular, we observe more preferable responses than binary labels and significant improvements where modestly-confident labels are in the majority. © 2024, CC BY.

关键词： optimization algorithms

Extremality of families of sets and set-valued optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Cuong, Nguyen Duy Kruger, Alexander Y. Thao, Nguyen Hieu Department of Mathematics College of Natural Sciences Can Tho University Can Tho City Viet Nam Optimization Research Group Faculty of Mathematics and Statistics Ton Duc Thang University Ho Chi Minh City Viet Nam School of Science Engineering and Technology RMIT University Vietnam Ho Chi Minh City Viet Nam

The paper continues our recent work [7] where another extension of the extremal principle has been established. We demonstrate its applicability to set-valued optimization problems with general preferences, weakening the assumptions of the known results and streamlining their proofs. Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms