检索结果-内蒙古大学图书馆

optimization over Trained Neural Networks: Difference-of-Convex Algorithm and Application to Data Center Scheduling

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Liu, Xinwei Dvorkin, Vladimir Department of Electrical Engineering and Computer Science University of Michigan MI48109 United States

When solving decision-making problems with mathematical optimization, some constraints or objectives may lack analytic expressions but can be approximated from data. When an approximation is made by neural networks, the underlying problem becomes optimization over trained neural networks. Despite recent improvements with cutting planes, relaxations, and heuristics, the problem remains difficult to solve in practice. We propose a new solution based on a bilinear problem reformulation that penalizes ReLU constraints in the objective function. This reformulation makes the problem amenable to efficient difference-of-convex algorithms (DCA), for which we propose a principled approach to penalty selection that facilitates convergence to stationary points of the original problem. We apply the DCA to the problem of the least-cost allocation of data center electricity demand in a power grid, reporting significant savings in congested cases. © 2025, CC BY-SA.

关键词： optimization algorithms

Preference optimization via Contrastive Divergence: Your Reward Model is Secretly an NLL Estimator

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Chen, Zhuotong Liu, Fang Zhu, Xuan Wang, Haozhu Qi, Yanjun Ghavamzadeh, Mohammad Amazon Web Services Santa ClaraCA United States SunnyvaleCA United States

Existing studies on preference optimization (PO) have centered on constructing pairwise preference data following simple heuristics, such as maximizing the margin between preferred and dispreferred completions based on human (or AI) ranked scores. However, none of these heuristics has a full theoretical justification. In this work, we develop a novel PO framework that provides theoretical guidance to effectively sample dispreferred completions. To achieve this, we formulate PO as minimizing the negative log-likelihood (NLL) of a probability model and propose to estimate its normalization constant via a sampling strategy. As we will demonstrate, these estimative samples can act as dispreferred completions in PO. We then select contrastive divergence (CD) as the sampling strategy, and propose a novel MC-PO algorithm that applies the Monte Carlo (MC) kernel from CD to sample hard negatives w.r.t. the parameterized reward model. Finally, we propose the OnMC-PO algorithm, an extension of MC-PO to the online setting. On popular alignment benchmarks, MC-PO outperforms existing SOTA baselines, and OnMC-PO leads to further improvement. © 2025, CC BY.

关键词： optimization algorithms

Visual Explanations of Some Problematic Search Behaviors of Frequently-Used EMO algorithms 13th

学校读者我要写书评

暂无评论

Visual Explanations of Some Problematic Search Behaviors of ...

13th International Conference on Evolutionary Multi-Criterion optimization, EMO 2025

作者： Ishibuchi, Hisao Pang, Lie Meng Guangdong Provincial Key Laboratory of Brain-Inspired Intelligent Computation Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen518055 China

ISBN: (纸本)9789819635375

Recently, evolutionary multi-objective optimization (EMO) algorithms have been used in various application fields. Whereas many new EMO algorithms are proposed every year, well-known EMO algorithms such as NSGA-II, MOEA/D, SMS-EMOA and NSGA-III have been still frequently used in application papers as shown by their number of citations. Since most EMO algorithms have been designed for and evaluated by artificial test problems such as DTLZ and WFG, they do not always work well on real-world problems. Main difficulties of real-world problems are as follows: (i) their Pareto front shapes are irregular, (ii) some objectives have much larger values than others, (iii) some decision variables have much larger values than others, (iv) some objectives are highly correlated, and (v) solution evaluation is expensive. In this paper, we focus on the first four difficulties (except for expensive solution evaluation). Our goal is to clearly explain problematic search behaviors of NSGA-II, MOEA/D, SMS-EMOA and NSGA-III caused by these difficulties using simple distance minimization problems. We formulate various distance minimization problems with the above-mentioned four difficulties by changing the location of each target point on the map as well as the size and shape of the map. We also suggest simple remedies for performance improvement of each algorithm for real-world applications without changing the algorithm itself. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： optimization algorithms

Multi-constraint optimization for Energy Scheduling of Plug-In Electric Vehicles 8th

学校读者我要写书评

暂无评论

Multi-constraint Optimization for Energy Scheduling of Plu...

8th International Conference on Life System Modeling and Simulation, LSMS 2024 and 8th International Conference on Intelligent Computing for Sustainable Energy and Environment, ICSEE 2024

作者： Zhang, Linxin Yang, Zhile Guo, Yuanjun Xiao, Qinge Zhu, Juncheng Faculty of Data Science City University of Macau Taipa999078 China Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong Shenzhen518055 China

ISBN: (纸本)9789819602315

The integration of a substantial number of plug-in electric vehicles (PEVs) into power grid scheduling introduces complexities due to stochastic charging and discharging behaviors, which pose significant challenges to the stability and reliability of the power system. The PEV scheduling optimization extends the traditional combinatorial optimization problem, adding constraints and variables, and becomes a multi-constraint unit commitment (MCUC)problem. In this paper, a novel PEV scheduling optimization framework is proposed to address MCUC problem, taking economic cost and carbon emissions as the objectives, a multi-constraint mathematical model is established. Meanwhile, a new binary level-based learning optimization algorithm is proposed to solve flexible scheduling of PEVs in power systems. Finally, considering different numbers of units, the superiority of the proposed optimization algorithm is verified by comparison with different algorithms. The feasibility and effectiveness of the proposed solution are also analyzed before and after scheduling to verify its potential in solving the MCUC problem. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： optimization algorithms

An Adaptive Dropout Approach for High-Dimensional Bayesian optimization

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Huang, Jundi Zhan, Dawei School of Computing and Artificial Intelligence Southwest Jiaotong University Chengdu China

Bayesian optimization (BO) is a widely used algorithm for solving expensive black-box optimization problems. However, its performance decreases significantly on high-dimensional problems due to the inherent high-dimensionality of the acquisition function. In the proposed algorithm, we adaptively dropout the variables of the acquisition function along the iterations. By gradually reducing the dimension of the acquisition function, the proposed approach has less and less difficulty to optimize the acquisition function. Numerical experiments demonstrate that AdaDropout effectively tackle high-dimensional challenges and improve solution quality where standard Bayesian optimization methods often struggle. Moreover, it achieves superior results when compared with state-of-the-art high-dimensional Bayesian optimization approaches. This work provides a simple yet efficient solution for high-dimensional expensive optimization. Copyright © 2025, The Authors. All rights reserved.

关键词： optimization algorithms

Performance-Driven optimization of Parallel Breadth-First Search

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Bhaskar, Marati Kanakagiri, Raghavendra Department of Computer Science and Engineering IIT Tirupati India

Breadth-first search (BFS) is a fundamental graph algorithm that presents significant challenges for parallel implementation due to irregular memory access patterns, load imbalance and synchronization overhead. In this paper, we introduce a set of optimization strategies for parallel BFS on multicore systems, including hybrid traversal, bitmap-based visited set, and a novel non-atomic distance update mechanism. We evaluate these optimizations across two different architectures - a 24-core Intel Xeon platform and a 128-core AMD EPYC system - using a diverse set of synthetic and real-world graphs. Our results demonstrate that the effectiveness of optimizations varies significantly based on graph characteristics and hardware architecture. For small-diameter graphs, our hybrid BFS implementation achieves speedups of 3 − 8× on the Intel platform and 3 − 10× on the AMD system compared to a conventional parallel BFS implementation. However, the performance of large-diameter graphs is more nuanced, with some of the optimizations showing varied performance across platforms including performance degradation in some cases. © 2025, CC BY.

关键词： optimization algorithms

An Optimal Algorithm for Half-plane Hitting Set 8

学校读者我要写书评

暂无评论

An Optimal Algorithm for Half-plane Hitting Set

8th SIAM Symposium on Simplicity of algorithms, SOSA 2025

作者： Liu, Gang Wang, Haitao Kahlert School of Computing University of Utah Salt Lake CityUT84112 United States

ISBN: (纸本)9781611978315

Given a set P of n points and a set H of n half-planes in the plane, we consider the problem of computing a smallest subset of points such that each half-plane contains at least one point of the subset. The previously best algorithm solves the problem in O(n3 log n) time. It is also known that Ω(n log n) is a lower bound for the problem under the algebraic decision tree model. In this paper, we present an O(n log n) time algorithm, which matches the lower bound and thus is optimal. Another virtue of the algorithm is that it is relatively simple. Copyright © 2025 by SIAM.

关键词： optimization algorithms

An Ensemble Niching Technique-Based Differential Evolution for Multimodal optimization Problems 19th

学校读者我要写书评

暂无评论

An Ensemble Niching Technique-Based Differential Evolution f...

19th CCF Conference on Computer Supported Cooperative Work and Social Computing, ChineseCSCW 2024

作者： Liang, Shao-Min Wang, Zi-Jia Chen, Zong-Gan Zhan, Zhi-Hui School of Computer Science and Cyber Engineering Guangzhou University Guangzhou510006 China School of Computer Science South China Normal University Guangzhou510898 China School of Computer Science and Engineering South China University of Technology Guangzhou510006 China

ISBN: (纸本)9789819623754

The multimodal optimization problems (MMOPs) require finding multiple optima simultaneously. Evolutionary computation integrated with niching techniques is commonly used to solve MMOPs. However, different niching techniques are suitable for different situations. In order to make the niche techniques perform better in different problems, in this paper, we propose an ensemble niching technique (ENT). In ENT, three different niching techniques, crowding clustering, speciation clustering, and nearest-better clustering (NBC), are used to identify the potential niche centers. Based on the potential niche center, the probability of an individual being a true niche center is calculated and the true niche center is identified. The identified true niche center can form niches more properly which combines the superiority of different niche techniques. Based on ENT, we combine it with DE and propose a novel multimodal optimization algorithm ENTDE. Meanwhile, we propose a dual-strategy update mechanism (DUM) to select the suitable mutation strategy in different evolutionary states for better evolution. To testify to the effectiveness of the proposed algorithm, ENTDE and eleven state-of-the-art algorithms are experimented on the widely used CEC2013 multimodal benchmark test suite. The experimental results indicate that ENTDE is significantly better than the compared algorithms. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： optimization algorithms

FABLE: A Bundle Method For Federated Learning In Wireless Systems

学校读者我要写书评

暂无评论

FABLE: A Bundle Method For Federated Learning In Wireless Sy...

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Cederberg, Daniel Larsson, Erik G. Johansson, Mikael Department of Electrical Engineering Stanford University Stanford United States Department of Electrical Engineering Linköping University Linköping Sweden Division of Decision and Control Systems KTH Royal Institute of Technology Stockholm Sweden

ISBN: (纸本)9798350368741

This paper presents a comprehensive approach to federated learning in wireless networks. We discuss communication strategies that address packet loss and bitrate limitations in both uplink and downlink transmissions, and introduce FABLE, a novel optimization algorithm designed to operate effectively under these network constraints. The algorithm also supports non-smooth regularizers and accommodates heterogeneous data distributions across clients. We provide theoretical convergence guarantees under gradient compression and asynchronous operation, and demonstrate the efficiency of our approach through numerical experiments. © 2025 IEEE.

关键词： Federated learning optimization algorithms wireless communications