Towards Explainable Evolution Strategies with Large Language Models

TL;DR

The paper tackles the opacity of Evolution Strategies (ES) by introducing Explainable Evolution Strategies (XES), which couples a self-adaptive $$(\mu + \lambda)$$-ES with a restart mechanism and a detailed logging system. An LLM is then used to translate rich optimization logs into concise narratives, highlighting convergence trends, optimal fitness values, and encounters with local optima. Through a case study on the 10-dimensional Rastrigin function, the authors demonstrate how multiple LLMs and prompting strategies can produce coherent explanations while enabling potential hyperparameter insights. The work emphasizes the feasibility of using LLM-generated explanations to bridge optimization algorithms and interpretability, and discusses limitations related to context length and the need for interactive, actionable summaries.

Abstract

This paper introduces an approach that integrates self-adaptive Evolution Strategies (ES) with Large Language Models (LLMs) to enhance the explainability of complex optimization processes. By employing a self-adaptive ES equipped with a restart mechanism, we effectively navigate the challenging landscapes of benchmark functions, capturing detailed logs of the optimization journey. The logs include fitness evolution, step-size adjustments and restart events due to stagnation. An LLM is then utilized to process these logs, generating concise, user-friendly summaries that highlight key aspects such as convergence behavior, optimal fitness achievements, and encounters with local optima. Our case study on the Rastrigin function demonstrates how our approach makes the complexities of ES optimization transparent. Our findings highlight the potential of using LLMs to bridge the gap between advanced optimization algorithms and their interpretability.