Quasidifferential optimization algorithms in a parallel computation environment

In this paper some parallel algorithms for the minimization of a quasidifferentiable function in the sense of Dem’yanov are considered. In particular a new parallel method for the search of a descent direction of a subdifferentiable function is presented. Such a method is based on the approximation of the subdifferential by a simplex which is related to the directional derivatives of the function at the current point; the direction of descent is found by solving in parallel some quadratic programming problems on the simplex. Some ideas about the possibility of reducing the number of constraints are also presented. Based on this new method, an algorithm for quasidifferentiable functions is sketched.