检索结果-内蒙古大学图书馆

Multitask Augmented Random Search in deep reinforcement learning

APPLIED SOFT COMPUTING 2024年 160卷

作者： Thanh, Le Tien Thang, Ta Bao Van Cuong, Le Binh, Huynh Thi Thanh Grad Univ Adv Studies Dept Informat SOKENDAI Tokyo Japan Hanoi Univ Sci & Technol Sch Informat & Commun Technol Hanoi Vietnam

Reinforcement Learning (RL) has gained significant popularity in recent years for its ability to solve complex control problems. However, most existing RL algorithms are designed to train policies for each environment in isolation, limiting their applicability to real -world scenarios with many related environments. Recently, many multitask optimization algorithms have been proposed and successfully applied to a wide range of optimization problems. However, existing studies mainly focus on solving multiple continuous function benchmarks and overlook the potential of tailoring towards RL. In this paper, we propose a simple multitask optimization algorithm called Multitask Augmented Random Search (MARS) that trains multiple RL agents together and exploits the performance surplus from highly correlated tasks. MARS is a modification of the simple random search Augmented Random Search (ARS) algorithm, which has been shown to outperform complicated methods in solving continuous control MuJoCo environments such as Soft Actor -Critic (SAC), Proximal Policy Optimization (PPO), and Trusted Region Policy Optimization (TRPO). The experimental results also demonstrate that our proposed algorithm is more consistent in solving different instances of MuJoCo benchmark than ARS, multifactorial Evolution algorithm (MFEA), and Adaptive MFEA RL (AMFEARL) within the same number of training episodes.

关键词： multifactorial evolutionary algorithm Deep reinforcement learning Augmented random search

来源：评论

学校读者我要写书评

暂无评论

Contrastive variational auto-encoder driven convergence guidance in evolutionary multitasking

引用

APPLIED SOFT COMPUTING 2024年 163卷

作者： Wang, Ruilin Feng, Xiang Yu, Huiqun Shanghai Engn Res Ctr Smart Energy Shanghai 200237 Peoples R China

Knowledge transfer is at the core of the evolutionary Multitasking (EMT) problem, as it exploits the interaction of inter-task common knowledge to accelerate task convergence. However, existing research on EMT is generally based solely on the framework of evolutionary computation, with limited integration of deep learning models. Additionally, few studies have found that utilizing deep learning models to generate individuals for transfer and guide the evolutionary trajectory may promote better convergence of algorithms. To address this research gap, we introduce the MFEA-VC (multifactorial evolutionary-Variational Auto-Encoder and Contrastive Learning) algorithm. Individuals are categorized based on task-label and inputted into a VAE, with sampling along feature dimensions. The VAE effectively guides the population towards better search areas by learning the latent trends of the distribution and generating transferred individuals. Simultaneously, a new training objective based on contrastive learning is proposed. This objective regulates the similarity between individuals from the same and different task-label, finely controlling individual features in the latent space. This approach makes the generated superior individuals more interpretable. To verify the superiority of MFEA-VC, we conduct comprehensive empirical studies on multi-task single-objective scenarios. We also validate the effectiveness of our improved loss function through theoretical analysis. The results demonstrate that, compared to state-of-the-art multifactorial algorithms, our method significantly enhances the global search capability during the early evolution stages, achieves excellent convergence results, and exhibits strong adaptability to heterogeneous tasks.

关键词： multifactorial evolutionary algorithm Contrastive learning Variational auto-encoder Knowledge transfer

来源：评论

学校读者我要写书评

暂无评论

Multi-objective evolutionary multi-tasking algorithm using cross-dimensional and prediction-based knowledge transfer

引用

INFORMATION SCIENCES 2022年 586卷 540-562页

作者： Chen, Qunjian Ma, Xiaoliang Yu, Yanan Sun, Yiwen Zhu, Zexuan Shenzhen Univ Coll Comp Sci & Software Engn Shenzhen Peoples R China BGI Shenzhen Shenzhen 518083 Peoples R China Shenzhen Univ Sch Biomed Engn Shenzhen Peoples R China

Transfer learning is an important research topic in machine learning and recently has been introduced into evolutionary computation to form evolutionary multi-task optimization (EMTO). EMTO focuses on tackling multiple optimization tasks simultaneously based on knowledge transfer and reuse. Many EMTO algorithms have been proposed mainly for single-objective optimization problems and achieved success in various fields, yet multi objective EMTO remains a big challenge. The existing multi-objective EMTO algorithms tend to suffer from issues like slow convergence and degraded performance on less correlated tasks. To alleviate these issues, this paper proposes a new multi-objective EMTO algorithm by introducing cross-dimensional variable search and prediction-based individual search for efficient knowledge transfer. The cross-dimensional variable search optimizes a decision variable using information collected from other variables. The prediction based individual search performs individual mapping where the offspring solutions and the corresponding parent solutions are symmetrized about the predicted population center to maintain the population diversity. The proposed algorithm is tested on benchmark problems and the experimental results demonstrate the effectiveness and efficiency of the algorithm.(c) 2021 Elsevier Inc. All rights reserved.

关键词： Multi-objective optimization Multi-objective multi-task optimization multifactorial evolutionary algorithm evolutionary multi-tasking

来源：评论

学校读者我要写书评

暂无评论

System-in-package design using multi-task memetic learning and optimization

引用

MEMETIC COMPUTING 2022年第1期14卷 45-59页

作者： Dai, Weijing Wang, Zhenkun Xue, Ke Southern Univ Sci & Technol Sch Syst Design & Intelligent Mfg Shenzhen Peoples R China Southern Univ Sci & Technol Dept Comp Sci & Engn Shenzhen Peoples R China

System-in-Package (SiP) is an advanced packaging technology and developing rapidly in semiconductor industry. Electronic modules of this package type are individual integrated systems for specific applications. Therefore, those modules are usually characterized by multiple encapsulated components and sophisticated internal structures. However, such complexity brings great challenges to package design. Traditional methods, like design of experiments, response surface analysis, are widely used in this field, but their effectiveness drops rapidly due to increasing complexity. In current scenarios, not only do the amount of design variables increases, but also the modules have diverse design tasks to satisfy. Thereby, package design for SiP modules is a multi-task optimization problem. To resolve this issue, we propose a multi-task memetic learning and optimization algorithm, in which multi-output Gaussian process model and multifactorial evolutionary algorithm are employed. In this work, knowledge transfer between different tasks is activated during both the surrogate modeling and model optimization procedures. Several variants of the proposed algorithm are tested, and their modeling accuracy and optimization efficiency were compared. This interdisciplinary study shows the benefits of the memetic knowledge transfer mechanism in improving modeling and optimizing efficacy in multi-task scenarios and presents a viable approach to achieve both automation and optimization for complicated packaging design.

关键词： System-in-packaging Optimal design Multi-task memetic learning and optimization multifactorial evolutionary algorithm Gaussian process

来源：评论

学校读者我要写书评

暂无评论

evolutionary Multitask Optimization: a Methodological Overview, Challenges, and Future Research Directions

引用

COGNITIVE COMPUTATION 2022年第3期14卷 927-954页

作者： Osaba, Eneko Del Ser, Javier Martinez, Aritz D. Hussain, Amir Basque Res & Technol Alliance BRTA TECNALIA Ed 700 Derio 48160 Spain Univ Basque Country UPV EHU Bilbao 48013 Spain Edinburgh Napier Univ Edinburgh Midlothian Scotland

In this work, we consider multitasking in the context of solving multiple optimization problems simultaneously by conducting a single search process. The principal goal when dealing with this scenario is to dynamically exploit the existing complementarities among the problems (tasks) being optimized, helping each other through the exchange of valuable knowledge. Additionally, the emerging paradigm of evolutionary multitasking tackles multitask optimization scenarios by using biologically inspired concepts drawn from swarm intelligence and evolutionary computation. The main purpose of this survey is to collect, organize, and critically examine the abundant literature published so far in evolutionary multitasking, with an emphasis on the methodological patterns followed when designing new algorithmic proposals in this area (namely, multifactorial optimization and multipopulation-based multitasking). We complement our critical analysis with an identification of challenges that remain open to date, along with promising research directions that can leverage the potential of biologically inspired algorithms for multitask optimization. Our discussions held throughout this manuscript are offered to the audience as a reference of the general trajectory followed by the community working in this field in recent times, as well as a self-contained entry point for newcomers and researchers interested to join this exciting research avenue.

关键词： Transfer optimization Multitasking optimization evolutionary multitasking multifactorial evolutionary algorithm Multi-population multitasking

来源：评论

学校读者我要写书评

暂无评论

Two levels approach based on multifactorial optimization to solve the clustered shortest path tree problem

引用

evolutionary INTELLIGENCE 2022年第1期15卷 185-213页

作者： Huynh Thi Thanh, Binh Pham Dinh, Thanh Hanoi Univ Sci & Technol Sch Informat & Commun Technol Hanoi Vietnam Taybac Univ Fac Nat Sci & Technol Son La Vietnam

The Clustered Shortest-Path Tree Problem (CluSPT) has a great meaning in theoretical research as well as a wide range of applications in everyday life, especially in the field of network optimization. Being able to solve the CluSPT will path the way for improving practical systems such as agricultural irrigation and product distribution. multifactorial evolutionary algorithm (MFEA) is a variant of evolutionary algorithm (EA) aiming at simultaneously solving multiple factorial tasks which can be diverse in types, dimensionalities and search spaces. The applications of MFEA have yet to be fully exploited, but the realm has recently attracted much interest from the research community, especially those who are working on evolutionary multitasking. Considering these characteristics, this paper describes a new approach using the MFEA for solving the CluSPT. The MFEA has two tasks: the first is solving the CluSPT problem and the second is solving a new problem which is decomposed from the CluSPT problem. The goal of the first task is finding the fittest solution as possible for the CluSPT while the goal of the second task is determining the best tree (w.r.t. cost minimization) which envelops all vertices of the problem. This paper also introduces mutation, crossover and population initialization operators for a proposed MFEA to solve the CluSPT. Each of these operators deals with individuals in two phases, one of which ensures that the resulted individual is a spanning tree and the other phase ensures that each of its clustered sub-graphs is also a spanning tree. As the MFEA had to deal with these two tasks, a decoding method was also created to allow communication between the tasks. To assess the effectiveness of the proposed algorithm and methods, the authors implemented them on both Euclidean and Non-Euclidean instances. Experiment results show that the proposed MFEA outperformed the existing MFEA algorithms on two-thirds instances with average improvement value of 47%.

关键词： multifactorial evolutionary algorithm Clustered shortest-path tree problem evolutionary algorithms Genetic algorithm multifactorial optimization

来源：评论

学校读者我要写书评

暂无评论

A hybrid and adaptive evolutionary approach for multitask optimization of post-disaster traveling salesman and repairman problems

引用

COMPUTERS & OPERATIONS RESEARCH 2024年 167卷

作者： Ban, Ha-Bang Binh, Huynh Thi Thanh Do, Tuan Anh Tran, Cong Dao Nguyen, Su Hanoi Univ Sci & Technol Sch Informat & Commun Technol Hanoi Vietnam RMIT Univ Melbourne Australia FPT Software AI Ctr Hanoi Vietnam

Traveling Salesman Problem (TSP) and Traveling Repairman Problem (TRP) have been studied due to their real -world applications. While the TSP minimizes the total time to travel to all customers, the TRP minimizes the total waiting time between a main depot and customers. Solving two problems simultaneously is shown to obtain promising results due to their high similarity in solution representation and overlap in search space. However, these two problems in the context of post -disaster have not yet been considered in line with the three following restrictions. First, the salesman cannot travel on several destroyed roads. Second, the salesman needs additional time to remove debris, which means the debris removal time is added to the travel cost. Finally, we do not consider each vertex equally because they have different characteristics depending on their population or vulnerability. Therefore, reaching a vertex with higher priority takes more benefit overall. To tackle these problems, we first formalize TSP and TRP in post -disaster scenarios as TSPPD and TRPPD, respectively, and then propose the effective metaheuristic, MFEA-TS, combining a multifactorial evolutionary algorithm (MFEA) and Tabu Search (TS) to solve them simultaneously and efficiently. In the proposed MFEATS, the MFEA can explore the search space well by transferring knowledge between two problems, while the TS can exploit good solutions in the search space. The proposed algorithm overcomes the drawbacks of the existing MFEA algorithms due to good exploitation capacity and prevention of revisiting previous solution spaces. We further conduct extensive experiments to verify the performance of our MFEA-TS with TRPPD and TSPPD. The empirical results show that the proposed algorithm can give high -quality solutions on a range of benchmarks, thus confirming the impressive efficiency of the proposed formulations and algorithm MFEA-TS.

关键词： Traveling salesman problem Traveling repairman problem Post-disaster scenarios multifactorial evolutionary algorithm Tabu search

来源：评论

学校读者我要写书评

暂无评论

Node depth Representation-based evolutionary Multitasking Optimization for Maximizing the Network Lifetime of Wireless Sensor Networks

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2024年 128卷

作者： Dao, Tran Cong Tam, Nguyen Thi Binh, Huynh Thi Thanh Hanoi Univ Sci & Technol Sch Informat & Commun Technol Hanoi Vietnam VNU Univ Sci Hanoi Vietnam FPT Software AI Ctr Hanoi Vietnam

Wireless Sensor Networks (WSNs) face challenges related to limited energy resources in sensor nodes, making network lifetime and energy consumption critical considerations. In this paper, we introduce a novel approach to extend network lifetime and reduce energy consumption in WSNs by optimizing network architecture selection. We explore various architecture options, including single-layer and multi-layer networks, to determine the most suitable configuration. Several studies have applied knowledge transfer in evolutionary multitasking optimization to maximize network lifetime, achieving state-of-the-art results for the problem. However, they still have some drawbacks, such as redundant representations, i.e., many genotypes represent the same phenotype, and invalid solution generation. To address these limitations, we propose a new evolutionary multitasking algorithm incorporating efficient encoding to represent tree-structure solutions (network architectures) to the problem. Additionally, we tailor new genetic operators, such as recombination and mutation, to meet the specific requirements of our proposed evolutionary multitasking algorithm. The standout feature of our proposal lies in its ability to consistently generate valid solutions for the problem, which significantly reduces redundancy within the genotype search space and effectively satisfies the problem constraints. Additionally, our multitask evolutionary algorithm facilitates the exploration of various network architectures, harnessing knowledge transfer to enhance optimal performance and reduce computation time compared to single-task methods. We conduct comprehensive experiments on diverse datasets and use statistical tests, e.g., the Wilcoxon signed rank test, to verify the performance of our proposed algorithms. The empirical results demonstrate that our proposed algorithm outperforms existing methods and achieves state-of-the-art results in terms of solution quality, convergence rate, and running ti

关键词： evolutionary multitasking optimization multifactorial evolutionary algorithm Node-depth encoding Network lifetime Wireless sensor network

来源：评论

学校读者我要写书评

暂无评论

evolutionary Constrained Multi-task Optimization: Benchmark Problems and Preliminary Results 22

Evolutionary Constrained Multi-task Optimization: Benchmark ...

引用

Genetic and evolutionary Computation Conference (GECCO)

作者： Li, Yanchi Gong, Wenyin Li, Shuijia Univ Geosci Sch Comp Sci China Wuhan Peoples R China

ISBN: (纸本)9781450392686

Multi-task optimization (MTO) aims to solve multiple tasks simultaneously. However, multi-task evolutionary algorithms (MTEAs) hardly consider the problem with constraints, while most optimization problems, in reality, are with constraints. This study presents a benchmark of constrained multi-task optimization problems (CMTOPs), modified from the CEC2017 competition on evolutionary multi-task optimization and CEC2017 competition on constrained real-parameter optimization. Moreover, this study attempts to solve CMTOPs by incorporating constraint handling techniques into MTEAs. Experimental results demonstrate the complexity of the proposed benchmark in the context of CMTOPs and present the requirements for handling constraints in MTEAs. Our preliminary exploration also reveals prospects for the development of evolutionary algorithms in the area of constrained multi-task optimization. The Matlab source code can be obtained from https://***/intLyc/CMTO-Benchmark.

关键词： Constrained multi-task optimization evolutionary multi-task optimization constrained optimization evolutionary transfer optimization multifactorial evolutionary algorithm

来源：评论

学校读者我要写书评

暂无评论

Enhancing evolutionary multitasking optimization by leveraging inter-task knowledge transfers and improved evolutionary operators

引用

KNOWLEDGE-BASED SYSTEMS 2023年 259卷

作者： Ma, Xiaoliang Xu, Mingyu Yu, Yanan Liu, Hongjie Wang, Yan Wang, Lei Qi, Yutao Xiong, Jian Shenzhen Univ Coll Comp Sci & Software Engn Shenzhen 518060 Peoples R China Chinese Acad Sci Shenzhen Inst Adv Technol SIAT Shenzhen 518055 Peoples R China Xidian Univ Sch Comp Sci & Technol Xian 710071 Peoples R China Southwestern Univ Finance & Econ Sch Business Adm Chengdu 610074 Peoples R China

It is inefficient and time-consuming to begin the search from scratch for each optimization task. evolutionary multitasking optimization (EMTO) handles multiple tasks simultaneously, aiming at im-proving the solving quality of every task via the evolutionary algorithm (EA) and inter-task knowledge transfer. Thus, suitable evolutionary operators and inter-task effective knowledge transfer are two key factors for the success of EMTO. As one of the representative EMTO algorithms, the multifactorial evolutionary algorithm (MFEA) has attracted a lot of attention. However, MFEA has suffered from the issue of premature convergence and negative knowledge transfer among tasks with a low correlation. To handle these issues, this article enhances MFEA with two proposed strategies, namely a carefully -designed opposition-based learning (OBL) strategy and a carefully-designed differential evolution (DE) strategy, named MFDE-OBL for short. Both the proposed OBL and the proposed DE contain an inter -task strategy and an intra-task strategy. To improve the effectiveness of knowledge transfer, the inter-task OBL strategy learns a linear subspace mapping among task/tasks' subpopulations to transfer different search scales among tasks, while the inter-task DE strategy uses genetic information from another task to improve the population diversity with different scales and directions. Besides, the intra-task generalized-opposite-point-based OBL is used to enhance the global search ability, while the intra-task DE strategy consists of two complementary DE strategies to maintain a good balance between exploitation and exploration. Finally, the proposed algorithm is tested on both single-objective and multi-objective multi-tasking test suites. Experimental results have shown the efficiency and effectiveness of the proposed algorithm compared with both classical and state-of-the-art algorithms.(c) 2022 Published by Elsevier B.V.

关键词： evolutionary multitasking optimization multifactorial evolutionary algorithm Inter -task knowledge transfer Opposition -based learning

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：