• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2007, Vol. 29 ›› Issue (6): 65-67.

• 论文 • 上一篇    下一篇

基于改进进化策略的CSCL任务分配自适应决策方法

王慧华 朱兆辉 吴克立   

  • 出版日期:2007-06-01 发布日期:2010-06-03

  • Online:2007-06-01 Published:2010-06-03

摘要:

本文针对在伸缩性很大的多用户多任务(MU-MT)网络协同学习环境中,如何获得稳定的总体效益最优化的任务分配问题,提出了一种基于改进进化策略的自适应决策方法。该方法在传统进化策略基础上,通过进化方向控制和自适应选择算子设计等改进,优化了收敛速度,并防止抖动和早熟。本算法的优点主要在于:运算规模能依据CSCL环境的变化自动调整,具有很强的环境自适应性,因而是在未知网络环境中提供稳健协同性能的有效解决方案。

关键词: CSCL 自适应性 进化策略 稳定性

Abstract:

In this paper, an adaptive decision method ,which focuses on how to get stable total benefit when solving the problems of task distribution in diverseMU-MT Internet environments, is proposed. This method improves the traditional evolution strategy for controlling the evolution direction and designingrational selection operators. These improvements can optimize the speed of convergence, and avoid dithering and premature convergence additionally. The main advantage of the algorithm is that the size of operation can be adjusted automatically according to the CSCL environment, so it has strong self-ad aptability to the environment and can be an efficient scheme to support cooperation stably in unknown Internet environmens.

Key words: CSCL, adaptability, evolution strategy (ES), stability