A neural network for temporal pattern recognition in the presence of noise

A three-level neural network is presented that is capable of recognizing temporal sequences in the presence of the noise. The organization of the network consists of a first-level tree in which activations propagate as new input terms are received. The second level produces output activations based on the activations in the first level and interconnection strengths from the first to the second level. The learning rules that involve the third teaching level are described. Simulation results indicate that the system learns very quickly and performs very reliably compared to a standard backpropagation network