| 摘要 |
PROBLEM TO BE SOLVED: To provide a circuit for identifying multidimensional analog input data in classes while automatically learning it, and discriminating and outputting the class to which the input data belongs using the learning result. SOLUTION: This pattern identification device is composed of a neural network of four layers and a circuit of its control part. The circuit receives analog detailed input data by a detail input layer 1, generates rough input data of a conceptual rough input layer 2 rough relative to the detail input layer 1, groups the input data by learning between the rough input layer 2 and an identification output layer 3 using the rough input data, notes pattern data in the groups as candidates, regards it as temporarily adapted pattern data one by one to reproduce reproduction data in an internal reproduction layer 4 and to compare it with the input data for determination, thus determining the identification class of adapted analog input. Rough information of the analog input data is extracted to find a subset for retrieval, and the input data is verified and identified by the reproduction data. COPYRIGHT: (C)2008,JPO&INPIT
|
