A policy learned using evolutionary algorithms (like genetic algorithms) rather than gradient-based optimization.