混沌理论、高级元搜索算法及其新式深度学习架构优化应用:综述

Fractals Pub Date : 2024-04-05 DOI:10.1142/s0218348x24300010
AKIF AKGUL, YELl̇Z KARACA, MUHAMMED ALI PALA, MURAT ERHAN ÇIMEN, ALI FUAT BOZ, MUSTAFA ZAHID YILDIZ
{"title":"混沌理论、高级元搜索算法及其新式深度学习架构优化应用:综述","authors":"AKIF AKGUL, YELl̇Z KARACA, MUHAMMED ALI PALA, MURAT ERHAN ÇIMEN, ALI FUAT BOZ, MUSTAFA ZAHID YILDIZ","doi":"10.1142/s0218348x24300010","DOIUrl":null,"url":null,"abstract":"<p>Metaheuristic techniques are capable of representing optimization frames with their specific theories as well as objective functions owing to their being adjustable and effective in various applications. Through the optimization of deep learning models, metaheuristic algorithms inspired by nature, imitating the behavior of living and non-living beings, have been used for about four decades to solve challenging, complex, and chaotic problems. These algorithms can be categorized as evolution-based, swarm-based, nature-based, human-based, hybrid, or chaos-based. Chaos theory, as a useful approach to understanding neural network optimization, has the basic idea of viewing the neural network optimization as a dynamical system in which the equation schemes are utilized from the space pertaining to learnable parameters, namely optimization trajectory, to itself, which enables the description of the evolution of the system by understanding the training behavior, which is to say the number of iterations over time. The examination of the recent studies reveals the importance of chaos theory, which is sensitive to initial conditions with randomness and dynamical properties that are principally emerging on the complex multimodal landscape. Chaotic optimization, in this regard, accelerates the speed of the algorithm while also enhancing the variety of movement patterns. The significance of hybrid algorithms developed through their applications in different domains concerning real-world phenomena and well-known benchmark problems in the literature is also evident. Metaheuristic optimization algorithms have also been applied to deep learning or deep neural networks (DNNs), a branch of machine learning. In this respect, the basic features of deep learning and DNNs and the extensive use of metaheuristic algorithms are overviewed and explained. Accordingly, the current review aims at providing new insights into the studies that deal with metaheuristic algorithms, hybrid-based metaheuristics, chaos-based metaheuristics as well as deep learning besides presenting recent information on the development of the essence of this branch of science with emerging opportunities, applicability-based optimization aspects and generation of well-informed decisions.</p>","PeriodicalId":501262,"journal":{"name":"Fractals","volume":"35 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-04-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"CHAOS THEORY, ADVANCED METAHEURISTIC ALGORITHMS AND THEIR NEWFANGLED DEEP LEARNING ARCHITECTURE OPTIMIZATION APPLICATIONS: A REVIEW\",\"authors\":\"AKIF AKGUL, YELl̇Z KARACA, MUHAMMED ALI PALA, MURAT ERHAN ÇIMEN, ALI FUAT BOZ, MUSTAFA ZAHID YILDIZ\",\"doi\":\"10.1142/s0218348x24300010\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>Metaheuristic techniques are capable of representing optimization frames with their specific theories as well as objective functions owing to their being adjustable and effective in various applications. Through the optimization of deep learning models, metaheuristic algorithms inspired by nature, imitating the behavior of living and non-living beings, have been used for about four decades to solve challenging, complex, and chaotic problems. These algorithms can be categorized as evolution-based, swarm-based, nature-based, human-based, hybrid, or chaos-based. Chaos theory, as a useful approach to understanding neural network optimization, has the basic idea of viewing the neural network optimization as a dynamical system in which the equation schemes are utilized from the space pertaining to learnable parameters, namely optimization trajectory, to itself, which enables the description of the evolution of the system by understanding the training behavior, which is to say the number of iterations over time. The examination of the recent studies reveals the importance of chaos theory, which is sensitive to initial conditions with randomness and dynamical properties that are principally emerging on the complex multimodal landscape. Chaotic optimization, in this regard, accelerates the speed of the algorithm while also enhancing the variety of movement patterns. The significance of hybrid algorithms developed through their applications in different domains concerning real-world phenomena and well-known benchmark problems in the literature is also evident. Metaheuristic optimization algorithms have also been applied to deep learning or deep neural networks (DNNs), a branch of machine learning. In this respect, the basic features of deep learning and DNNs and the extensive use of metaheuristic algorithms are overviewed and explained. Accordingly, the current review aims at providing new insights into the studies that deal with metaheuristic algorithms, hybrid-based metaheuristics, chaos-based metaheuristics as well as deep learning besides presenting recent information on the development of the essence of this branch of science with emerging opportunities, applicability-based optimization aspects and generation of well-informed decisions.</p>\",\"PeriodicalId\":501262,\"journal\":{\"name\":\"Fractals\",\"volume\":\"35 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-04-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Fractals\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1142/s0218348x24300010\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Fractals","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1142/s0218348x24300010","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

元启发式技术因其在各种应用中的可调整性和有效性,能够以其特定的理论和目标函数代表优化框架。通过深度学习模型的优化,元启发式算法从大自然中汲取灵感,模仿生物和非生物的行为,用于解决具有挑战性、复杂和混乱的问题已有近四十年的历史。这些算法可分为基于进化的算法、基于蜂群的算法、基于自然的算法、基于人类的算法、混合算法或基于混沌的算法。混沌理论作为理解神经网络优化的一种有用方法,其基本思想是将神经网络优化视为一个动态系统,在该系统中,方程方案被用于从与可学习参数(即优化轨迹)相关的空间到其自身,从而通过理解训练行为(即随时间变化的迭代次数)来描述系统的演化。对近期研究的审查显示了混沌理论的重要性,混沌理论对具有随机性和动态特性的初始条件非常敏感,而随机性和动态特性主要出现在复杂的多模态景观中。在这方面,混沌优化在加快算法速度的同时,也增加了运动模式的多样性。混合算法应用于不同领域,涉及现实世界现象和文献中著名的基准问题,其意义也是显而易见的。元启发式优化算法还被应用于机器学习的一个分支--深度学习或深度神经网络(DNN)。在这方面,本文概述并解释了深度学习和 DNN 的基本特征以及元启发式算法的广泛应用。因此,本综述除了介绍有关这一科学分支本质发展的最新信息外,还旨在为涉及元启发式算法、基于混合的元启发式算法、基于混沌的元启发式算法以及深度学习的研究提供新的见解,并介绍新出现的机遇、基于适用性的优化方面以及生成知情决策。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
CHAOS THEORY, ADVANCED METAHEURISTIC ALGORITHMS AND THEIR NEWFANGLED DEEP LEARNING ARCHITECTURE OPTIMIZATION APPLICATIONS: A REVIEW

Metaheuristic techniques are capable of representing optimization frames with their specific theories as well as objective functions owing to their being adjustable and effective in various applications. Through the optimization of deep learning models, metaheuristic algorithms inspired by nature, imitating the behavior of living and non-living beings, have been used for about four decades to solve challenging, complex, and chaotic problems. These algorithms can be categorized as evolution-based, swarm-based, nature-based, human-based, hybrid, or chaos-based. Chaos theory, as a useful approach to understanding neural network optimization, has the basic idea of viewing the neural network optimization as a dynamical system in which the equation schemes are utilized from the space pertaining to learnable parameters, namely optimization trajectory, to itself, which enables the description of the evolution of the system by understanding the training behavior, which is to say the number of iterations over time. The examination of the recent studies reveals the importance of chaos theory, which is sensitive to initial conditions with randomness and dynamical properties that are principally emerging on the complex multimodal landscape. Chaotic optimization, in this regard, accelerates the speed of the algorithm while also enhancing the variety of movement patterns. The significance of hybrid algorithms developed through their applications in different domains concerning real-world phenomena and well-known benchmark problems in the literature is also evident. Metaheuristic optimization algorithms have also been applied to deep learning or deep neural networks (DNNs), a branch of machine learning. In this respect, the basic features of deep learning and DNNs and the extensive use of metaheuristic algorithms are overviewed and explained. Accordingly, the current review aims at providing new insights into the studies that deal with metaheuristic algorithms, hybrid-based metaheuristics, chaos-based metaheuristics as well as deep learning besides presenting recent information on the development of the essence of this branch of science with emerging opportunities, applicability-based optimization aspects and generation of well-informed decisions.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信