Cheating for Problem Solving: A Genetic Algorithm with Social Interactions

TL;DR

This paper introduces a genetic algorithm variant incorporating social interactions modeled by game theory, aiming to enhance optimization performance and avoid local optima in complex problems.

Contribution

It presents a novel genetic algorithm that integrates social interactions based on game theory, demonstrating improved results on combinatorial optimization tasks.

Findings

Significant performance improvements on Knapsack problems.

Social interactions help avoid local optima.

Game theory models influence algorithm effectiveness.

Abstract

We propose a variation of the standard genetic algorithm that incorporates social interaction between the individuals in the population. Our goal is to understand the evolutionary role of social systems and its possible application as a non-genetic new step in evolutionary algorithms. In biological populations, ie animals, even human beings and microorganisms, social interactions often affect the fitness of individuals. It is conceivable that the perturbation of the fitness via social interactions is an evolutionary strategy to avoid trapping into local optimum, thus avoiding a fast convergence of the population. We model the social interactions according to Game Theory. The population is, therefore, composed by cooperator and defector individuals whose interactions produce payoffs according to well known game models (prisoner's dilemma, chicken game, and others). Our results on Knapsack problems show, for some game models, a significant performance improvement as compared to a standard genetic algorithm.