Title:
METHOD OF AUTONOMOUSLY CONSTITUTING FUZZY NEURAL NETWORK
Document Type and Number:
Japanese Patent JP2001142865
Kind Code:
A
Abstract:
To determine the initial location of membership functions corresponding to the non-linearity of a modeling object.
The calculation process of an inference value based on fuzzy inference is composed of a neural network, a parameter to be identified in the fuzzy inference is made correspondent to the coupling load of the neural network and the coupling load is updated by learning so that a fuzzy rule and a membership function can be controlled. In a learning process, the fuzzy rule is autonomously added and deleted and on a fuzzy neural network for acquiring the fuzzy rules of a number required for a controlled system, the number of initial membership functions is determined while using a predetermined evaluation reference.
Inventors:
UCHIKAWA YOSHIKI
FURUHASHI TAKESHI
YAMAGUCHI MASASHI
FUJIME YOKO
FURUHASHI TAKESHI
YAMAGUCHI MASASHI
FUJIME YOKO
Application Number:
JP32529399A
Publication Date:
May 25, 2001
Filing Date:
November 16, 1999
Export Citation:
Assignee:
UCHIKAWA YOSHIKI
FURUHASHI TAKESHI
YAMAHA MOTOR CO LTD
FURUHASHI TAKESHI
YAMAHA MOTOR CO LTD
International Classes:
G05B13/02; G06F9/44; G06F15/18; G06N3/00; G06N3/04; G06N7/02; (IPC1-7): G06F15/18; G05B13/02; G06F9/44
Attorney, Agent or Firm:
Hiroki Shirai (3 outside)