Cross entropy-based memetic algorithms: An application study over the tool switching problem

Abstract

This paper presents a parameterized schema for building memetic algorithms based on cross-entropy (CE) methods. This novel schema is general in nature, and features multiple probability mass functions and Lamarckian learning. The applicability of the approach is assessed by considering the Tool Switching Problem, a complex combinatorial problem in the field of Flexible Manufacturing Systems. An exhaustive evaluation (including techniques ranging from local search and evolutionary algorithms to constructive methods) provides evidence of the effectiveness of CE-based memetic algorithms.

Keywords

• Cross-entropy method
• memetic algorithms
• hybridization
• tool switching problem