Fast permutation learning

Abstract

Permutations occur in a great variety of optimization problems, such as routing, scheduling and assignment problems. The present paper introduces the use of learning automata for the online learning of good quality permutations. Several centralized and decentralized methods using individual and common rewards are presented. The performance, memory requirement and scalability of the presented methods is analyzed. Results on well known benchmark problems show interesting properties. It is also demonstrated how these techniques are successfully applied to multi-project scheduling problems. © 2012 Springer-Verlag.