On consistency of stationary points of stochastic optimization problems in a Banach space

Recently, Balaji and Xu studied the consistency of stationary points, in the sense of the Clarke generalized gradient, for the sample average approximations to a one-stage stochastic optimization problem in a separable Banach space with separable dual. We present an alternative approach, showing that the restrictive assumptions that the dual space is separable and the Clarke generalized gradient is a (norm) upper semicontinuous and compact-valued multifunction can be dropped. For that purpose, we use two results having independent interest: a strong law of large numbers and a multivalued Komlós theorem in the dual to a separable Banach space, and a result on the weak* closedness of the expectation of a random weak* compact convex set.