© 1967 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||
Minimizing a function without calculating derivatives
Center for Research in Management Science, University of California, Berekely, California, USA
In an important contribution Powell has suggested an approach for determining the unconstrained minimum of a function of several variables, and determining it without calculating derivatives. This paper studies his approach in some detail. It is first shown by counter-example that his basic method for minimizing a quadratic function in a finite number of iterations contains an error. His modification of his basic method is then simplified, and the simplification proven to converge for strictly convex functions. Finally, we pose a new method not only which converges in a finite number of iterations for a quadratic, but also for which theoretical convergence is established in the strictly convex case.
* Center for Research in Management Science, University of California, Berkeley, California, U.S.A.