Abstract
Co-evolution is a powerful problem-solving approach. However, fitness evaluation in co-evolutionary algorithms can be computationally expensive, as the quality of an individual in one population is often determined by its interactions with many (or all) members of another population. To accelerate co-evolutionary systems, we introduce phylogeny-informed interaction estimation, which uses runtime phylogenetic analysis to estimate interaction outcomes between individuals based on how their relatives performed against each other. In addition, we introduce a method for subsampling interactions which improves the accuracy of our estimates. We test our method on sorting networks using lexicase selection, and demonstrate that interaction estimation can halve the computation required to make progress in co-evolutionary search.