Abstract

A new architecture for optical implementation of large-scale neural networks is proposed. This architecture is based on a time-division-multiplexing technique, in which both the neuron state vector and the interconnection matrix are divided in the time domain. Computer simulation and experimental results for associative memories show the effectiveness in implementing large-scale networks.

© 1990 Optical Society of America

Full Article  |  PDF Article

References

  • View by:
  • |
  • |
  • |

  1. D. E. Rumelhart, J. L. McClellandthe PDP Research Group, Parallel Distributed Processing (MIT Press, Cambridge, Mass., 1986), Vols. 1 and 2.
  2. Y. S. Abu-Mostafa, D. Psaltis, Sci. Am. 256(3), 66 (1986).
  3. E. G. Paek, D. Psaltis, Opt. Eng. 26, 428 (1987).
  4. N. H. Farhat, D. Psaltis, A. Prata, E. Paek, Appl. Opt. 24, 1469 (1985).
    [CrossRef] [PubMed]
  5. J. Ohta, S. Tai, M. Oita, K. Kuroda, K. Kyuma, K. Hamanaka, Appl. Opt. 28, 2426 (1989).
    [CrossRef] [PubMed]
  6. H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
    [CrossRef]
  7. J. J. Hopfield, Proc. Natl. Acad. Sci. USA 79, 2554 (1982).
    [CrossRef] [PubMed]
  8. W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

1989 (1)

1987 (1)

E. G. Paek, D. Psaltis, Opt. Eng. 26, 428 (1987).

1986 (1)

Y. S. Abu-Mostafa, D. Psaltis, Sci. Am. 256(3), 66 (1986).

1985 (1)

1984 (1)

H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
[CrossRef]

1983 (1)

W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

1982 (1)

J. J. Hopfield, Proc. Natl. Acad. Sci. USA 79, 2554 (1982).
[CrossRef] [PubMed]

Abu-Mostafa, Y. S.

Y. S. Abu-Mostafa, D. Psaltis, Sci. Am. 256(3), 66 (1986).

Anderson, R.

W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

Caulfield, H. J.

H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
[CrossRef]

Farhat, N. H.

Hamanaka, K.

Hopfield, J. J.

J. J. Hopfield, Proc. Natl. Acad. Sci. USA 79, 2554 (1982).
[CrossRef] [PubMed]

Kuroda, K.

Kyuma, K.

McClelland, J. L.

D. E. Rumelhart, J. L. McClellandthe PDP Research Group, Parallel Distributed Processing (MIT Press, Cambridge, Mass., 1986), Vols. 1 and 2.

Ohta, J.

Oita, M.

Paek, E.

Paek, E. G.

E. G. Paek, D. Psaltis, Opt. Eng. 26, 428 (1987).

Prata, A.

Psaltis, D.

E. G. Paek, D. Psaltis, Opt. Eng. 26, 428 (1987).

Y. S. Abu-Mostafa, D. Psaltis, Sci. Am. 256(3), 66 (1986).

N. H. Farhat, D. Psaltis, A. Prata, E. Paek, Appl. Opt. 24, 1469 (1985).
[CrossRef] [PubMed]

W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

Ross, W. E.

W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

Rumelhart, D. E.

D. E. Rumelhart, J. L. McClellandthe PDP Research Group, Parallel Distributed Processing (MIT Press, Cambridge, Mass., 1986), Vols. 1 and 2.

Stermer, R. L.

H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
[CrossRef]

Tai, S.

Verber, C. M.

H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
[CrossRef]

Appl. Opt. (2)

Opt. Commun. (1)

H. J. Caulfield, C. M. Verber, R. L. Stermer, Opt. Commun. 51, 213 (1984).
[CrossRef]

Opt. Eng. (2)

E. G. Paek, D. Psaltis, Opt. Eng. 26, 428 (1987).

W. E. Ross, D. Psaltis, R. Anderson, Opt. Eng. 22, 485 (1983).

Proc. Natl. Acad. Sci. USA (1)

J. J. Hopfield, Proc. Natl. Acad. Sci. USA 79, 2554 (1982).
[CrossRef] [PubMed]

Sci. Am. (1)

Y. S. Abu-Mostafa, D. Psaltis, Sci. Am. 256(3), 66 (1986).

Other (1)

D. E. Rumelhart, J. L. McClellandthe PDP Research Group, Parallel Distributed Processing (MIT Press, Cambridge, Mass., 1986), Vols. 1 and 2.

Cited By

OSA participates in CrossRef's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (5)

Fig. 1
Fig. 1

Basic configuration of an optical neural network using the TDM technique.

Fig. 2
Fig. 2

Computer simulation results for the TDM architecture. The recognition rate is plotted as a function of the Hamming distance for various numbers of elements of the optical processors p.

Fig. 3
Fig. 3

Arrangement for the optical implementation of the associative memory using the TDM technique.

Fig. 4
Fig. 4

Eight stored vectors. The binary elements of the vectors are rearranged two dimensionally. The black and white squares represent one and zero, respectively.

Fig. 5
Fig. 5

Experimental and computer simulation results of the recognition rate for the same initial vectors using the TDM technique.

Equations (3)

Equations on this page are rendered with MathJax. Learn more.

υ I = υ t p + i , t = 0 , 1 ,…, N / p 1 ; i = 1 , 2 ,…, p ,
W I , J = W t p + i , T p + j , T = 0 , 1 ,…, N / p 1 ; j = 1 , 2 ,…, p ,
u T p + j = t = 0 time N / p 1 i = 1 space p W t p + i , T p + j υ t p + i

Metrics