论文信息 - Convergence Rates For The Distribution Of Program Outputs

Convergence Rates For The Distribution Of Program Outputs

Fitness distributions (landscapes) of programs tend to a limit as they get bigger. Markov chain convergence theorems give general upper bounds on the linear program sizes needed for convergence. Tight bounds (exponential in N, N log N and smaller) are given for five computer models (any, average, cyclic, bit flip and Boolean). Mutation randomizes a genetic algorithm population in 1/4 (l + 1)(log(l) + 4) generations. Results for a genetic programming (GP) like model are confirmed by experiment.

William B. Langdon | W. Langdon

[1] David H. Wolpert,et al. No free lunch theorems for optimization , 1997, IEEE Trans. Evol. Comput..

[2] Thomas Bäck,et al. Evolutionary algorithms in theory and practice - evolution strategies, evolutionary programming, genetic algorithms , 1996 .

[3] Riccardo Poli,et al. Foundations of Genetic Programming , 1999, Springer Berlin Heidelberg.

[4] Jeffrey S. Rosenthal,et al. Convergence Rates for Markov Chains , 1995, SIAM Rev..

[5] Tobias Blickle,et al. Theory of evolutionary algorithms and application to system synthesis , 1997 .

[6] Peter Nordin,et al. Genetic programming - An Introduction: On the Automatic Evolution of Computer Programs and Its Applications , 1998 .

[7] P. Diaconis. Group representations in probability and statistics , 1988 .

[8] Marc Schoenauer,et al. Rigorous Hitting Times for Binary Mutations , 1999, Evolutionary Computation.