Algoritmos Quasi-Newton: desempenho dos métodos BFGS, L-BFGS e DFP

Abstract

This work presents a comparative analysis of the quasi-Newton methods BFGS, L-BFGS, and DFP for Hessian approximation in numerical optimization problems. The study evaluates convergence rates, the average number of iterations, and the accuracy of the solutions obtained for different test functions. The methodology includes the implementation of the algorithms, computational experiments, and result analysis. The methods are compared in terms of efficiency and precision, highlighting their advantages and limitations. The findings contribute to the selection of an appropriate quasi-Newton method for optimization applications.