Theoretical Study of Optimizing Rugged Landscapes with the cGA

Estimation of distribution algorithms (EDAs) provide a distribution-based approach for optimization which adapts its probability during the run algorithm. We contribute to theoretical understanding EDAs and point out that their makes them more suitable deal with rugged fitness landscapes than classical local search algorithms. Concretely, we make OneMax function by adding noise each value. The cGA can nevertheless find solutions $$n(1-\varepsilon )$$ many 1s, even high variance noise. In contrast this, RLS (1+1) EA, probability, only $$n(1/2+o(1))$$ small variance.