Recombination and Novelty in Neuroevolution: A Visual Analysis

Abstract Neuroevolution has re-emerged as an active topic in the last few years. However, there is a lack of accessible tools to analyse, contrast and visualise behaviour neuroevolution systems. A variety search strategies have been proposed such Novelty Quality-Diversity search, but their impact on evolutionary dynamics not well understood. We propose using data-driven, graph-based model, trajectory networks (STNs) directly different methods. Our analysis uses NEAT for solving maze problems with two strategies: novelty-based fitness-based, including excluding crossover operator. model trajectories, contrasting illuminating studied variants. results confirm advantages novelty this setting, challenge usefulness recombination.