Técnicas de otimização para minimização unidimensional.

Abstract

In this paper a discussion of one-dimensional minimization methods is made. Several algorithms are described, from the simplest and most basic, to those more refined that deal with additional difficulties such as non-differentiability of the function to be minimized. Four hybrid algorithms are described for the minimization of functions continuously differentiable. A more detailed study of an algorithm for minimizing a class of functions that are not continuously differentiable has been done. Are presented, in a test report, the results of performance tests of hybrid algorithms with proteoses. Finally, a theoretical introduction is presented on the algorithms that search the minima along a curvilinear path, using directions of negative curvature.