Markov Chain Analyses of Random Local Search and Evolutionary Algorithm
Hiroshi Furutani, Hiroki Tagami, Makoto Sakamoto, Yifei Du · Journal of Robotics Networking and Artificial Life · 2014
Theoretical studies of evolutionary algorithms (EAs) have been developed by researchers whose main interests are convergence properties of algorithms.In this paper, we report the computational complexity of an algorithm that is a variant of (1 + 1) EA, called Random Local Search (RLS).While a standard EA uses a mutation of flipping each bit in a parent string, RLS flips exactly one bit at each step.It has been noted the close resemblance of RLS with the coupon collector problem (CCP).CCP has a long history of probabilistic research, and many interesting results are obtained.This study makes use of such results with some modifications.We also show some useful results representing the evolution process of (1 + 1) EA.