Improving CBR-LA algorithm to variable size problems

Abstract

In this paper an improved approach based on CBR-LA model is proposed for static task assignment in heterogeneous computing systems. The proposed model is composed of case based reasoning (CBR) and learning automata (LA) techniques. The LA is used as an adaptation mechanism that adapts previously experienced cases to the problem which must be solved (new case). The goal of this paper is to expressing some weak points of the CBR-LA and proposing new algorithm called ICBR-LA which has improved performance in terms of Makespan performance metric. The results of experiments have shown that the proposed model performs better than the previous one.