Title:
NEURAL NET AND ITS OPTIMIZING METHOD/ DEVICE
Document Type and Number:
Japanese Patent JPH0652134
Kind Code:
A
Abstract:
PURPOSE: To provide a neural net which is applicable for various optimizing problems and the optimizing method and device.
CONSTITUTION: Respective elements 1 divide the neural net of a mutual connection type, in which a saturation type linear function is set to be an input function, into plural blocks 2, synchronously update the state of the elements 1 in the blocks 2 and a synchronously update them for the respective blocks 2. Since convergence is ensured for an arbitrary and symmetrical synapse string W by selecting cut width for the respective problems in the neural net, the neural net can be applicable to the arbitrary optimizing problems.
More Like This:
Inventors:
SHIRATANI FUMIYUKI
Application Number:
JP20360592A
Publication Date:
February 25, 1994
Filing Date:
July 30, 1992
Export Citation:
Assignee:
OLYMPUS OPTICAL CO
International Classes:
G06F15/18; G06G7/60; G06N99/00; (IPC1-7): G06F15/18; G06G7/60
Attorney, Agent or Firm:
Ryuyoshi Abe (7 outside)
Next Patent: LEARNING SYSTEM FOR NEURAL NETWORK