Learning algorithm for pulse coupled neural network

Living neurons act as leaky integrators in that they store on their surface signals applied to them in previous fractions of a second. For large pyramidal neurons in the cortex the time constant of the cell membrane may be higher than had previously been thought and the temporal properties of the neurons need to be reassessed in this light. In any case, time constants of tens, or even hundreds of milliseconds lead to the storage of ongoing activity in a way that may be crucial in determining the possible modes of information processing. We propose a system in which the sequence is stored directly in a temporal manner. The system also leads to recall in which there is no distortion of the timing of the temporal patterns.