ARTIFICIAL NEURAL NETWORK SYSTEM USING PIECEWISE LINEAR RECTIFICATION UNIT FOR COMPENSATING FOR DEFECTS ON ELEMENT

摘要

An artificial neural network system of the present invention may comprise: an input layer for receiving input data; an output layer for outputting output data corresponding to the input data; and a hidden layer group including a plurality of hidden layers positioned between the input layer and the output layer, the hidden layers being connected to each other through connection weights, and the connection weights being updated on the basis of a difference between pieces of prediction data corresponding to the output data and the input data, wherein each of the hidden layers comprises a hidden node including an activation function, the activation function outputs "0" when an input value is negative and determines an output value on the basis of linear functions having different slopes according to the range of an input value when the input value is positive, and the slope of the linear functions decreases in a direction of going away from the origin.