Reinforcement learning algorithms as function optimizers

摘要

Any nonassociative reinforcement learning algorithm can be viewedas a method for performing function optimization through (possiblynoise-corrupted) sampling of function values. A description is given ofthe results of simulations in which the optima of several deterministicfunctions studied by D.H. Ackley (Ph.D. Diss., Carnegie-Mellon Univ.,1987) were sought using variants of REINFORCE algorithms. Resultsobtained for certain of these algorithms compare favorably to the bestresults found by Ackley