Uses the neural gas algorithm to construct a scenario tree for use in multi-stage stochastic programming. The primary input is a set of initial scenarios or realizations of a disturbance. The scenario tree nodal structure must be predefined using a scenario tree nodal partition matrix.