Towards user-centric memetic algorithms: Experiences with the TSP

Abstract

User-centric evolutionary computation is an optimization pa-radigm that tries to integrate the human user and the evolutionary algorithm in a smooth way, favoring bi-directional communication and establishing synergies among these two actors. We explore the possibilities for such an approach in the context of memetic algorithms, with application to the travelling salesman problem. Some ways to canalize this cooperation via the introduction of dynamic constraints and selective local search are hinted, and implementation and interfacing issues are discussed. The reported experiments on TSPLIB instances provide encouraging results for these techniques. © 2011 Springer-Verlag.