首页>
外国专利>
NEURAL NETWORK TYPE LEARNING DEVICE AND LEARNING DEGREE CALCULATING METHOD, NEURAL NETWORK TYPE LEARNING DEVICE, AND RECORDING MEDIUM RECORDED WITH LEARNING DEGREE CALCULATING PROGRAM
NEURAL NETWORK TYPE LEARNING DEVICE AND LEARNING DEGREE CALCULATING METHOD, NEURAL NETWORK TYPE LEARNING DEVICE, AND RECORDING MEDIUM RECORDED WITH LEARNING DEGREE CALCULATING PROGRAM
PROBLEM TO BE SOLVED: To easily calculate the learning capacity of a neural network type learning machine and the degree of learning at the end of the learning. ;SOLUTION: A fixed value and characteristic vector calculation part 4 calculates fixed values and characteristic vectors of a differential matrix of 2nd order of an error function after learning by a learning part 3 consisting of a multi-layered artificial neural network and a learning degree calculation part 5 calculates the degree of learning from the fixed values and characteristic vectors. An output part 6 outputs the degree of learning found by the learning degree calculation part 5 from the learning result of the learning part 3. The degree of learning can be calculated only by finding the fixed values and characteristic values of the differential matrix of 2nd order of the error function at the end of a normal learning process. This is effectively usable for a pet robot which needs to perform an independent learning process, an independent moving robot, and a device for global optimizing solution, etc.;COPYRIGHT: (C)1999,JPO
展开▼