Title:
NEURAL NETWORK AND ITS LEARNING METHOD
Document Type and Number:
Japanese Patent JP3169635
Kind Code:
B2
Abstract:
PURPOSE: To provide the neural network which can learn even a linearly unseparable pattern and the learning method for that by conquering the problem that a perceptron can not linearly be separated and the problem that the local minimum point of back propagation is reached.
CONSTITUTION: In the neural network which has an input layer 1, an intermediate layer 2, and an output layer 3, a neural network model of low order consists of neurons of the intermediate layer 2 and input layer 1 which learn plural linearly separable patterns and a neural network model of high order consists of neurons of the intermediate layer 2 and output layer 3 which provide linearly unseparable patterns by integrating those linear separable patterns.
Inventors:
Ryujin Watanabe
Application Number:
JP16314291A
Publication Date:
May 28, 2001
Filing Date:
July 03, 1991
Export Citation:
Assignee:
Honda motor industry stock company
International Classes:
G06G7/60; G06N3/00; G06F15/18; G06N3/04; G06N99/00; (IPC1-7): G06N3/00; G06G7/60
Domestic Patent References:
| JP3125257A | ||||
| JP2285462A |
Attorney, Agent or Firm:
Tatsuhiko Sato (3 others)