PURPOSE: To process the information at a high speed through a neural network by performing an operation to obtain the output value of the neuron of an s-th layer and an operation to obtain the output value of the neuron of an (s + 1)-th layer with at least a partial duplication secured between both operations.
CONSTITUTION: Two operations are carried out to obtain the output value of the neuron of an s-th layer and that of an (s + 1)-th layer with at least a partial duplication secured between both operations. Thus an operation is performed to obtain the output value of an i-th neuron of the c-th layer based on the output values of all neurons of an (s - 1)-th layer and the product of weight coefficients obtained between the neuron of the (s - 1)-th layer and the i-th neuron of the s-th layer. At the same time, the product of weight coefficients is calculated between the neuron of the (s + 1)-th layer and an (i - 1)-th neuron of the s-th layer together with the output value of the (i -1 )-th neuron of the s-th layer. Thus the information can be processed at a high speed through a neural network.
KIMURA KATSUTAKA
ITO KIYOO
