ELM-Based Agents for Grid Resource Selection
Guopeng Zhao, Zhiqi Shen, Ailiya, Chunyan Miao · 2008
Resource selection in Grid involves great dynamics and uncertainties inherited from tasks and resources. The optimal selection of a resource against a task requires fast and intelligent services. Intelligent agent with fast learning capability is promising to resource selection problem in Grid. This paper proposes an Extreme Learning Machine (ELM)-based agent, in which an ELM connectionist module is embedded in an extended Belief-Desire-Intention (BDI) architecture. ELM empowers the agent with fast training and learning speed in the Grid environment. To improve generalization performance a cooperative learning among a group of ELM-based agents is proposed, for which the group decision is summarized upon individual decisions. The experiment results show that ELM-based agents are able to provide intelligent resource selection services, and the proposed cooperative learning outperforms the individual one.