Deep memetic models for combinatorial optimization problems: application to the tool switching problem

TL;DR

This paper introduces deep memetic models that leverage multi-level cooperative optimization algorithms to improve solutions for the Tool Switching Problem, demonstrating superior performance over existing metaheuristics.

Contribution

It presents a novel deep meta-cooperative framework for memetic algorithms, analyzing structural parameters and validating its effectiveness on a combinatorial problem.

Findings

Deep models outperform traditional metaheuristics on the Tool Switching Problem.

Structural parameters like communication topology influence model performance.

Deep cooperation enhances solution quality in combinatorial optimization.

Abstract

Memetic algorithms are techniques that orchestrate the interplay between population-based and trajectory-based algorithmic components. In particular, some memetic models can be regarded under this broad interpretation as a group of autonomous basic optimization algorithms that interact among them in a cooperative way in order to deal with a specific optimization problem, aiming to obtain better results than the algorithms that constitute it separately. Going one step beyond this traditional view of cooperative optimization algorithms, this work tackles deep meta-cooperation, namely the use of cooperative optimization algorithms in which some components can in turn be cooperative methods themselves, thus exhibiting a deep algorithmic architecture. The objective of this paper is to demonstrate that such models can be considered as an efficient alternative to other traditional forms of cooperative algorithms. To validate this claim, different structural parameters, such as the communication topology between the agents, or the parameter that influences the depth of the cooperative effort (the depth of meta-cooperation), have been analyzed. To do this, a comparison with the state-of-the-art cooperative methods to solve a specific combinatorial problem, the Tool Switching Problem, has been performed. Results show that deep models are effective to solve this problem, outperforming metaheuristics proposed in the literature.