Learning Rate Variable and Error Signal Size Limitation Method

摘要

The present invention relates to a learning rate variable and an error signal size limitation method for learning a neural network, and more particularly, to a neural network widely used for learning a pattern recognition problem, an error which occurs when a generalized cross-entropy error function is used for efficient learning By providing a method for limiting the magnitude of the error signal for the variable learning rate and the weight change as a method for solving the sensitivity of the learning function to the degree of the function, the learning performance is improved by the error function in the learning using the generalized cross- It is not greatly influenced.