Fitness Uniform Selection to Preserve Genetic Diversity

Title: Fitness Uniform Selection to Preserve Genetic Diversity

Abstract: This research proposes a new selection scheme for evolutionary algorithms called Fitness Uniform Selection (FUSS). Unlike standard selection schemes that favor individuals with higher fitness, FUSS generates selection pressure towards sparsely populated fitness regions. This approach can be much more effective than standard selection schemes in finding the globally optimal solution.

Main Research Question: How can we design an effective selection scheme for evolutionary algorithms that preserves genetic diversity and does not necessarily favor higher fitness individuals?

Methodology: The study uses an evolutionary algorithm with a population of individuals, each with a fitness value. The FUSS scheme selects individuals uniformly in the fitness values, meaning it generates selection pressure towards sparsely populated fitness regions, not necessarily towards higher fitness. This approach is compared to standard selection schemes like proportionate, truncation, rank, and tournament selection.

Results: The research shows that FUSS can be more effective than standard selection schemes in finding the globally optimal solution. This is because FUSS preserves genetic diversity better, allowing the algorithm to escape from local optima and explore other fitness regions.

Implications: The FUSS scheme can be applied to various optimization problems where standard selection schemes may not be effective. This can lead to improved performance and better solutions. The research also contributes to the broader field of evolutionary algorithms by introducing a new selection scheme that balances the trade-off between exploration and exploitation.

Link to Article: https://arxiv.org/abs/0103015v1 Authors: arXiv ID: 0103015v1