PURPOSE: To increase the calculation speed by providing plural ring networks having the functions of an input layer, an intermediate layer, and an output layer and executing the learning function of the back propagation method for given patterns in parallel by respective ring networks.

CONSTITUTION: This device has the multilayered structure consisting of an input layer, at least one intermediate layer, and an output layer, and plural ring networks A, B, C, and D having respective functions of the input layer, intermediate layers, and the output layer execute the learning function of the back propagation method for patterns given to these ring networks in parallel. Thus, the calculation speed is increased by parallel calculation of plural elemental processors, and plural patterns are simultaneously learnt in the simulation of a neural network to considerably increase the calculation speed.