Numerical experiment with new conjugate direction methods for nondifferentiable optimization

Results obtained by using a new algorithm for nondifferentiable optimization are presented. This algorithm is an extension of the Wolfe-Lemarechal algorithm, which simulates conjugate direction methods for differentiable problems. The author tested two versions of his algorithm. The first one is intended for nondifferentiable problems whose whole subdifferential is easily determined. The second one requires only one subgradient of the functional at the given point. The results reported should be treated qualitatively. The author presents the results of the steepest descent method, the Wolfe-Lemarechal algorithm, and the new algorithm.<
>