Optical neural network processing element with multiple holographic element interconnects

• US 5,129,041 A
• Filed: 06/08/1990
• Issued: 07/07/1992
• Est. Priority Date: 06/08/1990
• Status: Expired due to Term

• Current Assignee: Grumman Aerospace Corporation (Northrop Grumman Corporation)
• Original Assignee: Grumman Aerospace Corporation (Northrop Grumman Corporation)
• Inventors: Leib, Kenneth G., Brandstetter, Robert W., Pernick, Benjamin J.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/535,152
• Time in Patent Office: 760 Days
• Field of Search: 364/513, 364/807, 364/822, 364/841, 350/3.68, 395/25
• US Class Current: 706/40
• CPC Class Codes: G06N 3/0675 using electro-optical, acousto-optical or opto-electronic means

Abstract

A neural network processing element uses primarily optical components to model a biological neuron having both spatial and temporal dependence. The neural network processing element includes a switch-controlled laser source, a multiple holographic lens, a spatial/temporal light modulator, and a photodetector array. Laser beam control may be optical, electrical or acoustical, or a combination of these.

Claims

1. An optical neural network processing element comprising:

source means for forming an input coherent light beam;

multiple holographic lens means for receiving said input light beam and outputting a light beam array;

modulation means for spatially modulating the output array of said multiple holographic lens means, and for temporally modulating the output array by varying a magnitude of light beams in the array over time; and

photodetector means for receiving and converting the optical output of said modulation means into electrical signals representative of said modulated output.

2. An output neural network processing element as claimed in claim 1, wherein said source means comprises:

a laser beam source;

summing means for summing signals input to said neural network processing element to produce a summed output signal; and

beam controller means arranged to receive said summed signal and to modulate the output of said laser beam source in accordance with said summed signal.

3. An optical neural network processing element as claimed in claim 2, further comprising means for electrically controlling said beam controller means.

4. An optical neural network processing element as claimed in claim 2, further comprising means for optically controlling said beam controller means.

5. An optical neural network processing element as claimed in claim 2, wherein said signals input to said neural network processing element are optical signals.

6. An optical neural network processing element as claimed in claim 2, wherein said beam controller means comprises means for temporally modulating said laser beam source output.

7. An optical neural network processing element as claimed in claim 2, wherein said beam controller means comprises a Fabry-Perot interferometer for modulating said laser beam source output by controlling relative mirror spacing of said Fabry-Perot interferometer.

8. An optical neural network processing element as claimed in claim 2, wherein said beam controller means comprises a Michelson interferometer for modulating said laser beam source output by translating at least one mirror of said Michelson interferometer.

9. An optical neural network processing element as claimed in claim 1, wherein said multiple holographic lens means establishes a place of sharp focus and excludes extraneous light beams that do not form part of said output array of said multiple holographic lens means.

10. An optical neural network processing element as claimed in claim 1, further comprising a laser diode array for receiving said electrical signals output from said photodetector means.

11. An optical neural network processing element comprising:

source means for forming an input coherent light beam;

multiple holographic lens means for receiving said input light beam and outputting a light beam array;

modulation means for spatially modulating the output array of said multiple holographic lens means, and for temporally modulating the output array by varying a polarization of light beams in the array over time; and

photodetector means for receiving and converting the optical output of said modulation means into electrical signals representative of said modulated output.

12. An optical neural network processing element as claimed in claim 11, wherein said multiple holographic lens means establishes a plane of sharp focus and excludes extraneous light beams that do not form part of said output array of said multiple holographic lens means.

13. An optical neural network processing element as claimed in claim 11, further comprising a laser diode array for receiving said electrical signals output from said photodetector means.

14. An optical neural network processing element as claimed in claim 11, wherein said source means comprises:

a laser beam source;

a summing means for summing signals input to said neural network processing element to produce a summed output signal; and

beam controller means arranged to receive said summed signal and to modulate the output of said laser beam source in accordance with said summed signal.

15. An optical neural network processing element as claimed in claim 11, further comprising means for electrically controlling said beam controller means.

16. An optical neural network processing element as claimed in claim 11, further comprising means for optically controlling said beam controller means.

17. An optical neural network processing element as claimed in claim 11, wherein said signals input to said neural network processing element are optical signals.

18. An optical neural network processing element as claimed in claim 11, wherein said beam controller means comprises means for temporally modulating said laser beam source output.

19. An optical neural network processing element as claimed in claim 11, wherein said beam controller means comprises a Fabry-Perot interferometer for modulating said laser beam source output by controlling relative mirror spacing of said Fabry-Perot interferometer.

20. An optical neural network processing element as claimed in claim 11, wherein said beam controller means comprises a Michelson interferometer for modulating said laser beam source output by translating at least one mirror of said Michelson interferometer.

21. An optical neural network processing element comprising:

source means for forming an input coherent light beam;

multiple holographic lens means for receiving said input light beam and outputting a light beam array;

modulation means for spatially modulating the output array of said multiple holographic lens means, and for temporally modulating the output array by varying the phase of light beams in the array over time; and

photodetector means for receiving and converting the optical output of said modulation means into electrical signals representative of said modulated output.

22. An optical neural network processing element as claimed in claim 21 wherein said multiple holographic lens means establishes a plane of sharp focus and excludes extraneous light beams that do not form part of said output array of said multiple holographic lens means.

23. An optical neural network processing element as claimed in claim 21, further comprising a laser diode array for receiving said electrical signals output from said photodetector means.

24. An optical neural network processing element as claimed in claim 21, wherein said source means comprises:

a laser beam source;

a summing means for summing signals input to said neural network processing element to produce a summed output signal; and

beam controller means arranged to receive said summed signal and to modulate the output of said laser beam source in accordance with said summed signal.

25. An optical neural network processing element as claimed in claim 21, further comprising means for electrically controlling said beam controller means.

26. An optical neural network processing element as claimed in claim 21, further comprising means for optically controlling said beam controller means.

27. An optical neural network processing element as claimed in claim 21, wherein said signals input to said neural network processing element are optical signals.

28. An optical neural network processing element as claimed in claim 21, wherein said beam controller means comprises means for temporally modulating said laser beam source output.

29. An optical neural network processing element as claimed in claim 21, wherein said beam controller means comprises a Fabry-Perot interferometer for modulating said laser beam source output by controlling relative mirror spacing of said Fabry-Perot interferometer.

30. An optical neural network processing element as claimed in claim 21, wherein said beam controller means comprises a Michelson interferometer for modulating said laser beam source output by translating at least one mirror of said Michelson interferometer.