PURPOSE: To increase the learning procedure speed by correcting a synapse coefficient in an electronic neuron circuit only at the time when the whole of a learning base is inputted by the input of a neuron network and all output vectors are calculated by the neuron network.

CONSTITUTION: An additional electronic neuron circuit is included in an electronic circuit for the purpose of calculating correction of the synapse coefficient, and the synapse coefficient is corrected by the electronic neuron circuit only when the whole of the learning base is inputted by the input of the network and all output vectors are calculated by the network. Preferably, calculation of correction for the synapse coefficient is performed in accordance with procedures obtained by an inverse propagation algorithm. Thus, the learning procedure speed is practically increased, and a quick learning method is realized.