Abstract
An evaluation is made of a proposed adaptive step-size random search method to minimize a quadratic function, and the computational results are replicated and improved. Comparing this method to the Fletcher-Powell technique, it is shown that the required number of function evaluations grow linearly with dimension for both algorithms, with the Fletcher-Powell method superior to the proposed random search technique. In addition, several interesting characteristics of the Fletcher-Powell method are experimentally noted.

This publication has 4 references indexed in Scilit: