Escaping Local Optima Using Crossover With Emergent Diversity.

Population diversity is essential for avoiding premature convergence in genetic algorithms (GAs) and for the effective use of crossover. Yet the dynamics of how diversity emerges in populations are not well understood. We use rigorous runtime analysis to gain insight into population dynamics and GA performance for the (μ + 1) GA and the Jump test function. We show that the interplay of crossover f...