Parallel minimax search for a maximum

Abstract

We consider the problem of locating within a specified tolerance the point at which a unimodal function of one variable assumes its maximum value. Parallel search policies are considered in which a sequence of k-tuples of function evaluations is made, with the selection of each k-tuple dependent on the outcomes of all previous function evaluations. Search policies for the maximum are obtained which are optimal in the minimax sense. © 1968 Academic Press Inc.