The paper presents a comparison between two different flavors of nonlinear models to be used for the approximate solution of T-stage stochastic optimization (TSO) problems, a typical paradigm of Markovian decision processes. Specifically, the well-known class of neural networks is compared with...