Propagation and control of stochastic signals through universal learning networks