PURPOSE: To provide the system for making the learning of a neural network more secure.
CONSTITUTION: A neural network NW is provided with plural output nodes N to show a specified judged result so that each output node N can respectively output time sequential data to an input, and the total of time sequential data outputted from the output nodes N is calculated by successively adding those data for each output node N. The maximum value of the output total values of the respective nodes N is selected, when the maximum value is larger than a threshold value decided in advance, the node N corresponding to the maximum value is defined as a recognized result and when the maximum value is smaller than the threshold value decided in advance, recognition is disabled.
JPH08124085 | ASPECT DESIGN DEVICE |
JPH10187651 | LEARNING TYPE RECOGNITION AND JUDGEMENT DEVICE |
JPH0312789 | ASSOCIATIVE CIRCUIT NETWORK |
KAKIMOTO MASANORI