Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
NEURAL NETWORK LEARNING DEVICE
Document Type and Number:
Japanese Patent JPH0561847
Kind Code:
A
Abstract:

PURPOSE: To provide a neural network learning device which can improve the high speed characteristic of learning, convergence for an optimum solution and safety for system-down.

CONSTITUTION: A storage device 6 storing a learning pattern, storage devices 4 and 5 storing a load value, storage elements 2 and 3 inputting and storing more than one learning pattern and outputting the whole components of a vector in one learning pattern in parallel and a neural network device 1 altering a load value held in a self internal part in accordance with the inputted learning pattern are provided. The learning pattern is transferred from the storage device 6 to the storage elements 2 and 3, and the learning pattern which the storage elements 2 and 3 output is set to be the input of the neural network device 1. The initial value of the load value held in the neural network device 1 is set to be the load value transferred from the storage device 4 or 5, and the load value altered in the neural network device 1 is transferred to the storage device 5 so as to store it.


Inventors:
KAMAYA YUKIO
Application Number:
JP22084491A
Publication Date:
March 12, 1993
Filing Date:
September 02, 1991
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
TOSHIBA CORP
International Classes:
G06F15/18; G06G7/60; G06N3/00; G06N99/00; (IPC1-7): G06F15/18; G06G7/60
Attorney, Agent or Firm:
Noriyuki Noriyuki