Abstract

Synapses are critical components of an artificial neural network. Bacteriorhodopsin thin film can be used to construct compact, finely graded synapses for an optoelectronic neural network, based on its photochromic properties. Measurements show that these photochromic changes are blocked at low temperature. Thermal gating will allow synapses to be written on the film optically and then read without erasure by the same light and also will permit local implementation of an associative learning rule for updating synaptic weights.

© 1997 Optical Society of America

Full Article  |  PDF Article

References

  • View by:
  • |
  • |
  • |

  1. M. Balberg, M. Razvag, S. Vidro, E. Rafaeli, and A. J. Agranat, Opt. Lett. 21, 1544 (1996).
    [CrossRef] [PubMed]
  2. K. Wagner and D. Psaltis, eds., feature on optical implementation of neural networks, Appl. Opt. 32, 1261–1476 (1993).
    [CrossRef] [PubMed]
  3. D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
    [CrossRef]
  4. D. Haronian and A. Lewis, Appl. Opt. 30, 97 (1991).
    [CrossRef]
  5. H. Takei, A. Lewis, Z. Chen, and I. Nebenzahl, Appl. Opt. 30, 500 (1991).
    [CrossRef] [PubMed]
  6. H. Takei and N. Shimizu, Opt. Lett. 21, 537 (1996).
    [CrossRef] [PubMed]
  7. R. Thoma, N. Hampp, C. Brachle, and D. Oesterhelt, Opt. Lett. 16, 651 (1991).
    [CrossRef] [PubMed]
  8. Q. W. Song, C. Zhang, R. Blumer, R. B. Gross, Z. Chen, and R. R. Birge, Opt. Lett. 18, 1373 (1993).
    [CrossRef]
  9. D. Oesterhelt, C. Brauchle, and N. Hampp, Q. Rev. Biophys. 24, 425 (1991).
    [CrossRef] [PubMed]
  10. R. R. Birge, Annu. Rev. Phys. Chem. 41, 683 (1990).
    [CrossRef]
  11. J. K. Lanyi, in Molecular and Biomolecular Electronics, R. R. Birge, ed., Vol.  240 of Advances in Chemistry (American Chemical Society, Washington, D.C., 1994), Chap.  20, p. 491.
    [CrossRef]
  12. S. P. Balashov and F. F. Litvin, Biophysics 25, 566 (1981).
  13. M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
    [CrossRef]
  14. S. Haykin, Neural Networks (Macmillan, Englewood Cliffs, N.J., 1994).
  15. T. H. Brown and S. Chattrji, in Models of Neural Networks II, E. Domany, J. L. van Hemmen, and K. Schulten, eds. (Springer, New York, 1994), Chap.  8, p. 287.
    [CrossRef]

1996 (3)

1993 (2)

1991 (4)

1990 (2)

R. R. Birge, Annu. Rev. Phys. Chem. 41, 683 (1990).
[CrossRef]

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

1981 (1)

S. P. Balashov and F. F. Litvin, Biophysics 25, 566 (1981).

Agranat, A. J.

Bacon, M.

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Balashov, S. P.

S. P. Balashov and F. F. Litvin, Biophysics 25, 566 (1981).

Balberg, M.

Baxter, R. L.

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Birge, R. R.

Blumer, R.

Brachle, C.

Brady, D.

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

Brauchle, C.

D. Oesterhelt, C. Brauchle, and N. Hampp, Q. Rev. Biophys. 24, 425 (1991).
[CrossRef] [PubMed]

Brown, T. H.

T. H. Brown and S. Chattrji, in Models of Neural Networks II, E. Domany, J. L. van Hemmen, and K. Schulten, eds. (Springer, New York, 1994), Chap.  8, p. 287.
[CrossRef]

Chattrji, S.

T. H. Brown and S. Chattrji, in Models of Neural Networks II, E. Domany, J. L. van Hemmen, and K. Schulten, eds. (Springer, New York, 1994), Chap.  8, p. 287.
[CrossRef]

Chen, Z.

Gross, R. B.

Gu, X. G.

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

Hampp, N.

Haronian, D.

D. Haronian and A. Lewis, Appl. Opt. 30, 97 (1991).
[CrossRef]

Haykin, S.

S. Haykin, Neural Networks (Macmillan, Englewood Cliffs, N.J., 1994).

Kar, A. K.

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Lanyi, J. K.

J. K. Lanyi, in Molecular and Biomolecular Electronics, R. R. Birge, ed., Vol.  240 of Advances in Chemistry (American Chemical Society, Washington, D.C., 1994), Chap.  20, p. 491.
[CrossRef]

Lewis, A.

Lin, S.

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

Litvin, F. F.

S. P. Balashov and F. F. Litvin, Biophysics 25, 566 (1981).

Nebenzahl, I.

Oesterhelt, D.

Psaltis, D.

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

Rafaeli, E.

Razvag, M.

Shimizu, N.

Song, Q. W.

Takei, H.

Thoma, R.

Vidro, S.

Wang, C. H.

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Wherett, B. S.

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Zhang, C.

Annu. Rev. Phys. Chem. (1)

R. R. Birge, Annu. Rev. Phys. Chem. 41, 683 (1990).
[CrossRef]

Appl. Opt. (3)

Biophysics (1)

S. P. Balashov and F. F. Litvin, Biophysics 25, 566 (1981).

Nature (London) (1)

D. Psaltis, D. Brady, X. G. Gu, and S. Lin, Nature (London) 343, 325 (1990).
[CrossRef]

Opt. Commun. (1)

M. Bacon, C. H. Wang, A. K. Kar, R. L. Baxter, and B. S. Wherett, Opt. Commun. 124, 175 (1996).
[CrossRef]

Opt. Lett. (4)

Q. Rev. Biophys. (1)

D. Oesterhelt, C. Brauchle, and N. Hampp, Q. Rev. Biophys. 24, 425 (1991).
[CrossRef] [PubMed]

Other (3)

J. K. Lanyi, in Molecular and Biomolecular Electronics, R. R. Birge, ed., Vol.  240 of Advances in Chemistry (American Chemical Society, Washington, D.C., 1994), Chap.  20, p. 491.
[CrossRef]

S. Haykin, Neural Networks (Macmillan, Englewood Cliffs, N.J., 1994).

T. H. Brown and S. Chattrji, in Models of Neural Networks II, E. Domany, J. L. van Hemmen, and K. Schulten, eds. (Springer, New York, 1994), Chap.  8, p. 287.
[CrossRef]

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 (6)

Fig. 1
Fig. 1

Photocycle of light-adapted BR with photoreactions shown as solid lines and thermally activated reactions shown as dashed lines. The spectroscopically distinguishable intermediate states K, L, N, and O are short-lived compared with M. The key states in the BR photocycle for this application are B, K, and M.

Fig. 2
Fig. 2

Absorption spectra of BR film at 77  K in the B and M states.

Fig. 3
Fig. 3

Schematic of the experimental apparatus. Pump (solid lines) and probe (dashed lines) light beams intersect in the BR sample.

Fig. 4
Fig. 4

Normalized B-state absorbance change versus temperature measured for wild-type BR/polymer film. Filled and open circles show measurements of the change in the absorbance peak for the B state owing to forward BM and reverse MB photoreactions, respectively. The fitted curve is the function FT=1+expT-170/15-1.

Fig. 5
Fig. 5

Schematics of (a) a generic neuron and (b) an optoelectronic neuron, with optical inputs xi, synaptic weights wi, neuron electrical activation Uiwixi, and neuron optical output fU. The synapses are implemented with a BR film (BR) and a photodiode (PIN), whose output goes to an integrated circuit (VLSI), which drives a laser (VCSEL). The dashed line indicates a path for control signals from the neuron back to the overlying synaptic layer.

Fig. 6
Fig. 6

Synaptic weight evolution calculated for a model neuron when the activation of a subset of the synapses, Ncorr in number, is correlated. The synaptic input intensity is random and binary, xi=0 or xi=1, with average ON time of 10% for all synapses. The synapses in the correlated subset are activated only simultaneously. The time evolution of the weights for synapses with correlated and uncorrelated inputs (solid and dashed curves, respectively), calculated with the mean-field approximation for cases with 0, 10, and 20 synchronously activated synapses, is shown. Only the synapses with correlated activity are reinforced. The model for the ith synapse is wi=expfi-1A0 ln 10, dfi/dt=xi1-fi-bfiFT, and FT=1+exp-T-170/15-1, where A0=2.0, xi and b are the yellow- and blue-light intensities, respectively, at the synapse, and fi=Mi/(Bi+Mi) is the fraction of photobleached BR molecules. The model for the neuron is U=giwixi, T=200U+77, and b=U, where the neuron gain is g=0.10 and there are 100 synapses.

Metrics