Neural network structure having lateral interconnections

• US 5,581,660 A
• Filed: 03/10/1994
• Issued: 12/03/1996
• Est. Priority Date: 03/16/1993
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Horan, Paul
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/208,328
• Time in Patent Office: 999 Days
• Field of Search: 395/20-27, 382/153-159, 382/107, 359/1-10, 359/107, 359/245, 359/306, 359/310, 364/807, 364/513, 364/822
• US Class Current: 706/40
• CPC Class Codes: G06N 3/0675 using electro-optical, acousto-optical or opto-electronic means

Abstract

A neural network layer (1) is made up of nodes or neurons which each comprise a pair of physically separate and optically coupled sub-units (X₁, Y₁). One sub-unit broadcasts excitatory and receives inhibitory signals, whereas the other sub-unit broadcasts inhibitory and receives the excitatory signals. An electrical feedback connection (20) is provided between corresponding sub-units for determination of net node activation. Diffractive or holographic optical elements (2) are used for optical coupling.

Claims

1. A lateral neural network structure comprising:

a layer having a plurality of neurons, each of said plurality of neurons having at least two physically separate optical sub-units, one constructed for broadcasting lateral inhibitory signals and one constructed for broadcasting lateral excitatory signals; and

means for optically coupling said sub-units in a bi-directional manner for broadcasting both inhibitory and excitatory lateral signals;

wherein sub-units constructed for broadcasting lateral inhibitory signals are located for reception of lateral excitatory signals and reception of external inhibitory inputs; and

sub-units constructed for broadcasting lateral excitatory are located for reception of lateral inhibitory signals and reception of external excitatory inputs.

2. A neural network structure as claimed in claim 1, wherein the network structure layer comprises a feedback connection between sub-units.

3. A neural network structure as claimed in claim 2, wherein the feedback connection is a one-to-one connection between sub-units of a single neuron.

4. A neural network structure as claimed in claim 3, wherein the sub-units comprise on-chip light sources.

5. A neural network structure as claimed in claim 3, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

6. A neural network structure as claimed in claim 2, wherein the sub-units comprise on-chip light sources.

7. A neural network structure as claimed in claim 2, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

8. A neural network structure as claimed in claim 7, wherein the on-chip modulators are of the GaAs asymmetric Fabry-Perot type.

9. A neural network structure as claimed in claim 7, wherein the external source is modulated.

10. A neural network structure as claimed in claim 1, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

11. A neural network structure as claimed in claim 10, wherein the on-chip modulators are of the GaAs asymmetric Fabry-Perot type.

12. A neural network structure as claimed in claim 10, wherein the external source is modulated.

13. A neural network structure as claimed in claim 11, wherein the optical coupling means is variable with time.

14. A neural network structure as claimed in claim 1, wherein the optical coupling means comprises holographic or diffractive optical elements.

15. A neural network structure as claimed in claim 14, wherein the optical coupling means comprises a computer-generated holographic pattern.

16. A neural network structure as claimed in claim 14, wherein the optical coupling means is variable with time.

17. A neural network structure as claimed in claim 1, wherein the optical coupling means is variable with time.

18. An image processing system incorporating a neural network structure as claimed in claim 1.

19. A self-organising network incorporating a neural network structure as claimed in claim 1.

20. A neural network structure as claimed in claim 1, wherein the network structure layer comprises a feedback connection between sub-units.

21. A neural network structure as claimed in claim 20, wherein the feedback connection is a one-to-one connection between sub-units of a single neuron.

22. A neural network structure as claimed in claim 21, wherein the sub-units comprise on-chip light sources.

23. A neural network structure as claimed in claim 21, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

24. A neural network structure as claimed in claim 21, wherein the optical coupling means comprises holographic or diffractive optical elements.

25. A neural network structure as claimed in claim 21, wherein the optical coupling means is variable with time.

26. A neural network structure as claimed in claim 20, wherein the sub-units comprise on-chip light sources.

27. A neural network structure as claimed in claim 20, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

28. A neural network structure as claimed in claim 27, wherein the on-chip modulators are of the GaAs asymmetric Fabry-Perot type.

29. A neural network structure as claimed in claim 27, wherein the external source is modulated.

30. A neural network structure as claimed in claim 20, wherein the optical coupling means comprises holographic or diffractive optical elements.

31. A neural network structure as claimed in claim 20, wherein the optical coupling means is variable with time.

32. A neural network structure as claimed in claim 1, wherein the sub-units comprise on-chip light sources.

33. A neural network structure as claimed in claim 1, wherein the sub-units comprise on-chip light sources.

34. A neural network structure as claimed in claim 33, wherein the optical coupling means comprises holographic or diffractive optical elements.

35. A neural network structure as claimed in claim 33, wherein the optical coupling means is variable with time.

36. A neural network structure as claimed in claim 1, wherein the sub-units are constructed to receive light from an external source and comprise on-chip modulators for the received light.

37. A neural network structure as claimed in claim 36, wherein the on-chip modulators are of the GaAs asymmetric Fabry-Perot type.

38. A neural network structure as claimed in claim 37, wherein the optical coupling means comprises holographic or diffractive optical elements.

39. A neural network structure as claimed in claim 36, wherein the external source is modulated.

40. A neural network structure as claimed in claim 39, wherein the optical coupling means comprises holographic or diffractive optical elements.

41. A neural network structure as claimed in claim 39, wherein the optical coupling means is variable with time.

42. A neural network structure as claimed in claim 36, wherein the optical coupling means comprises holographic or diffractive optical elements.

43. A neural network structure as claimed in claim 1, wherein the optical coupling means is variable with time.

44. A neural network structure as claimed in claim 36, wherein the optical coupling means is variable with time.