NEURAL NETWORK EVALUATION DEVICE AND NEURAL NETWORK EVALUATION METHOD

摘要

PROBLEM TO BE SOLVED: To calculate the probability that a neural network precisely recognizes a data belonging to a prescribed category. SOLUTION: An evaluation data input means 2 gains an evaluation data 1 classified every category and inputs it to the neural network 3. The neural network 3 performs an arithmetic processing to the inputted evaluation data. An output evaluation value gaining means 4 gains the data before passing the sigmoid function of the unit of the output layer of the neural network 3 as an output evaluation value. An output value judgment reference value gaining means 6 gains the threshold of the unit of the neural network 3 corresponding to the category to be evaluated as an output value judgment reference value. A recognition rate calculating means 5 calculates the recognition rate to a certain category by performing a statistic processing to the output evaluation value belonging to the certain category and the output value judgment reference value.