Learning and anticipation in online dynamic optimization with evolutionary algorithms: the stochastic case