摘要 |
PURPOSE:To correctly recognize a pattern over a long period in correspondence to a change with the lapse of time in the characteristic of a sensor group by changing the weight value of a neural net so that teaching data in the case of the recognition can close to a correspondent teaching data value at the time of initial learning based on the recognized result of the pattern recognition in a prescribed period. CONSTITUTION:A control means 60 receives an instruction from a timing instructing means 50 and changes the weight value of the neural net so that the teaching data of a synaps element 230 in such a case can close to the correspondent teaching data value at the time of the initial learning based on the recognized result of a recognizing means 30 in a case that the instruction is given. When one week passes after the change, the timing instructing means 50 instructs again relearning. With receiving the instruction, the control means 60 changes the weight again. Thus, in correspondence to the change with the lapse of time in the characteristic of the sensor group, the pattern can be correctly recognized over the long period. |