摘要
提出了一种基于分布式Agent的机器学习模型。用Agent来构成机器学习的各组成部分,并由采用不同学习方法的Agent构成学习模块。为实现分布式Agent的协作学习机制,采用Agent联盟作为协作体系结构。将一种改进的遗传算法用于Agent联盟以实现分布式Agent的协作,该遗传算法控制参数动态地变化,采用粗粒度型的并行方式,一传多的交换个体方式和完全网络拓扑迁移模型。系统具有良好的性能。
The paper proposed a machine-learning model based on distributed agents. The system comprised several agents, and agents with different study methods composed the learning part. To realize the cooperative study, the agent league was adopted as cooperative structure, and an improved inheritance algorithm was used in agent league to realize the cooperation of distributed agents. The heritance algorithm controlled the parameters dynamically, and adopted the collateral mode of coarse granularity. The one to multi-ones mode to exchange individuals and the all-connect-graph transference model were also adopted. The system performed very well.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2005年第6期79-81,共3页
Journal of Wuhan University of Technology
基金
湖北省科技攻关项目 (2 0 0 2AA10 2B0 1)
关键词
AGENT
机器学习
协调
协作联盟
遣传算法
Agent
machine learning
coordination
cooperation league
heritance algorithm
