On a neural network with synaptic weight having double minimum potential

The author proposes a generalized Hebb-type learning equation in which the connection weight of the synapse has an arbitrary potential function. He examines the case when the potential has double stable points and finds the learning process in pattern recognition and the ordering process in pattern formation. The neural network with synaptic weight having double stable points has the merit that the system keeps learned patterns stored after input signals are stopped.<>