A modified univariate search algorithm

This paper describes a modified univariate search algorithm that overcomes two major limitations of conventional univariate search method. It minimizes the probability of premature convergence to poor local minima by utilizing a non-deterministic search procedure based on an analogy with the analytical univariate search, and improves´ the quality of solutions by dealing with populations of solutions rather than with single solutions for solving unconstrained as well as constrained optimization problems involving continuous or discrete variables. Unlike Genetic Algorithms (GA´s), which also are based on non-deterministic search and exhibit intrinsic parallelism, the solutions do not interact or mix together to produce new solutions (offspring); instead, new solutions are generated by unilaterally updating a single variable at a time in individual solutions. Results of two test problems are presented and compared with those obtained by standard GA, modified GA, and an optimization program based on the method of feasible directions