【発明の名称】階層型ニューラルネットワークを用いたセンサ情報処理方式

Japanese Patent JP3112750

B2

PURPOSE:To solve an optimization problem wherein the best solution for restriction conditions showing the properties and state of an object is found for sensor data by using the hierarchic neural network in the sensor information processing system which recognizes the state of an outside field with information from a sensor. CONSTITUTION:Sensor information processing consists of a learning process 11 wherein data as a recognition result is supplied to an input layer unit and the sensor information regarding a body is supplied as tutor data to an output layer unit and a recognizing process 12 wherein a solution which satisfies the restriction conditions regarding the outside field with small contradiction between the output of the output layer unit and the sensor information is searched for by varying the input to the input layer unit and minimizing an energy function corresponding to the contradiction and restriction conditions, and this system is equipped with a means 13 which supplies input data corresponding to the prediction of the recognition result as an initial value to the input layer unit in the recognizing process 12.

JP2520331	PROBLEM TO BE SOLVED: To learn a general network.
JP2001022727	METHOD AND DEVICE FOR CLASSIFYING AND LEARNING TEXT AND STORAGE MEDIUM STORING TEXT CLASSIFYING AND LEARNING PROGRAM
JPH06314266	MODELING SUPPORTING DEVICE AND METHOD THEREFOR

Daiki Masumoto
Takashi Kimoto
Shigemi Nagata
Yoshinori Yaginuma

JP24624492A

November 27, 2000

September 16, 1992

Click for automatic bibliography generation  

富士通株式会社

G06F15/18; G01D21/02; G06G7/60; G06N3/00; G06N3/04; G06N99/00; G06T7/00; (IPC1-7): G06T7/00; G06G7/60; G06N3/00; G06N3/04

Yoshiyuki Osuge (1 outside)

Previous Patent: ペット用おむつカバー

Next Patent: ステッパーの構造