A Memetic Algorithm with self-adaptive local search: TSP as a case study

A promising hybridization scheme for a Memetic Algorithm composed of two optimization processes, a Genetic Algorithm and a Monte Carlo method, which comment it successful application to a molecular conformational problem: Protein Folding.

Abstract

In this paper we introduce a promising hybridization scheme for a Memetic Algorithm (MA). Our MA is composed of two optimization processes, a Genetic Algorithm and a Monte Carlo method (MC). In contrast with other GA-Monte Carlo hybridized memetic algorithms, in our work the MC stage serves two purposes: -- when the population is diverse it acts like a local search procedure and -- when the population converges its goal is to diversify the search. To achieve this, the MC is self-adaptive based on observations from the underlying GA behavior; the GA controls the long-term optimization process. We present preliminary, yet statistically significant, results on the application of this approach to the TSP problem.We also comment it successful application to a molecular conformational problem: Protein Folding.