Title:
NEURAL NETWORK DEVICE
Document Type and Number:
Japanese Patent JP3224831
Kind Code:
B2
Abstract:
PURPOSE: To offer a neural network device by which plural different learning methods whose constitutions are the same can be arbitrarily selected, without preparing plural kinds of hardwares (integrated circuits) in a neuron unit.
CONSTITUTION: Plural neuron units 4a, 4b and synapse units 5a, 5b which apply synapse loads to be updated based on a learning signal to input signals to the pertinent neuron units, are used as basic constituting elements. The plural learning signals corresponding to the different learning methods are prepared in the neuron units 4a, 4b. and one learning signal for updating the synapse load is selected from those learning signals by a learning signal selecting switch 27.
Inventors:
Ken Shima
Application Number:
JP22073791A
Publication Date:
November 05, 2001
Filing Date:
August 30, 1991
Export Citation:
Assignee:
Toshiba Corporation
International Classes:
G06F15/18; G06G7/60; G06N3/04; G06N3/063; G06N99/00; (IPC1-7): G06N3/063; G06G7/60
Domestic Patent References:
| JP2287860A | ||||
| JP3125257A |
Attorney, Agent or Firm:
Takehiko Suzue