This invention comprises a novel artificial network for learning, recognizing, and generating temporal (time-dependent) spacial processing that offers to improve simulations of natural and biologic systems including human learning processes. Previously developed artificial neural networks have limitations because they often do not recognize sequences for which they have not been pre-programmed and have great difficulty discriminating temporal spacial patterns. Furthermore, they have trouble processing images that are obscured by "noise." This newly developed system overcomes such limitations by incorporating time-delay signal circuits, comparator units, and a parallel array of subneural networks. It is capable of learning temporal-spacial sequences such as speech patterns, robotic and unmanned defense system control commands, and forecasts of multivariate stochastic processes (i.e., weather, stock market, etc.). The system is able to recall an entire sequence after being presented with only a small portion of the sequence, which may be obscured by noise and/or contain blank spacial patterns.