On nonsmooth and discontinuous problems of stochastic systems optimization

A group of nonsmooth stochastic optimization problems, such as risk-control, optimization of stochastic networks, and discrete event systems, were analyzed to demonstrate the pertinence of random search models in resolving such problems. The solutions to nonsmooth optimization problems cannot be found through deterministic and standard stochastic computation methods. Results show that the use of the stochastic mollifier gradient can produce approximation methods for systems with Lipschitz and discontinuous objective functions.