Computer Science · Artificial Intelligence
A Comparison of Genetic Algorithms using Super Mario Bros.
Karl Jansen Kuhn, Ross Patrick Foley
See on the mapAbstract
This project was designed to compare and contrast Evolving Behavior Trees (EBTs) with NeuroEvolution of Augmenting Topologies (NEAT), a Genetic Algorithm for the evolution of Artificial Neural Networks. We used Super Mario Bros. as a benchmark to compare these two techniques. The results showed that NEAT had a slightly higher maximum fitness while performing poorly in all other comparisons. EBTs performed strongly in rise time, evolution time, generalization, and complexity.