Stochastic generalized gradient method for nonconvex nonsmooth stochastic optimization (Q1280941)

The stochastic quasigradient method is generalized to minimization problems with a nonconvex set of feasible solutions and with nonconvex, nondifferentiable objective functions, possibly in the form of an expectation of a function dependent on random parameters. Generalized differentiability of the objective and of the constraints is assumed. The convergence result proved in the paper implies that limit points of the infinite sequences generated by the generalized stochastic quasigradient method fulfil the necessary optimality condition for the underlying problem.