项目名称: 基于机器学习技术的差分演化算法研究
项目编号: No.61305085
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 蔡奕侨
作者单位: 华侨大学
项目金额: 25万元
中文摘要: 当前大部分差分演化算法都忽略了迭代过程中产生的大量中间数据和结果。如何对这些信息进行挖掘和利用,对提升差分演化算法的性能具有非常重要的意义。因此,本项目拟将机器学习技术与差分演化算法进行结合,重点对以下三个方面进行系统研究:(1)研究基于当前种群进化信息的学习策略;(2)通过机器学习技术与学习策略的有效结合,研究积木块的发现和利用机制、搜索引导策略和种群管理方法;(3)从算法的搜索机制和收敛特性出发,分析基于机器学习技术的差分演化算法的搜索行为。研究旨在利用机器学习技术建立较完善的进化信息挖掘和利用框架,充分发挥进化信息引导算法搜索过程的作用。项目所得到的成果一方面能够改进差分演化算法的性能,为其他进化算法在挖掘和利用进化信息方面提供示例作用,另一方面能够从理论上指导机器学习技术与差分演化算法的协作,为基于机器学习技术的进化算法设计提供理论参考。
中文关键词: 差分演化算法;进化信息;机器学习技术;学习策略;数值优化
英文摘要: As an emerging computational intelligent method, differential evolution (DE) has been applied in many fields successfully. However, a great deal of data and results are produced and always neglected during the evolutionary process of the most DE algorithms. It is significance in improving the performance of DE that how to mine and exploit these evolutionary information. Therefore, in this project, we will combine the machine learning (ML) techniques and DE to study the effective way of mining and exploiting the evlutionary information. In this project, we will focus on the following key issues:(1) how to design the effective learning strategies based on the evolutionary information so as to improve the communication between individuals; (2) how to use ML in DE to mine and exploit the evlutionary information, with the learning strategies of individuals; (3) how to make the deeper theoretical analysis on the ML based DE algorithm to clarify its search behavior and the collaboration between ML and DE. Through the above research, this project aims to use ML to build a relatively perfect framework of mining and exploiting the evolutionary information . The results of our research will further improve the performance of DE and provide the examples for other evolutionary algorithms in terms of mining and exploiting
英文关键词: Differential evolution;evolutionary information;machine learning technique;learning strategy;numerical optimization