Diffraction neural networks based on 4F circuits with liquid crystal light modulators

Skidanov, R.V., Khanenko Y.V., Morozov A.E., Pronin A.S., Sorokin D.M.

For Russian citation (Opticheskii Zhurnal):

Скиданов Р.В., Ханенко Ю.В., Морозов А.Е., Пронин А.С., Сорокин Д.М. Дифракционные нейронные сети на основе 4F- схемы с жидкокристаллическими модуляторами света // Оптический журнал. 2025. Т. 92. № 12. С. 66–76. http://doi.org/10.17586/1023-5086-2025-92-12-66-76

Skidanov R.V., Khanenko Yu.V., Morozov A.E., Pronin A.S., Sorokin D.M. Diffraction neural networks based on 4F circuits with liquid crystal light modulators [in Russian] // Opticheskii Journal. 2025. V. 12. № 12. P. 66–76. http://doi.org/10.17586/1023-5086-2025-92-12-66-76

For citation (Journal of Optical Technology):

Abstract:

The subject of the study. The main factors influencing the error of the optical implementation of mathematical transformations in the 4F scheme are considered. The optical implementation of a convolutional neural network for classification problems is considered. The aim of the work is to confirm the possibility of a mixed implementation of neural networks from optical and electronic layers. Method. The work used both mathematical modeling of the 4F system and a full-scale experiment in which the 4F system was assembled as a compact device. The main results. It is shown that the main sources of errors in the 4F system are information input/output devices. Due to their imperfection, the error rate is too high for direct implementation of mathematical calculations. But using a compact neural layer of only 10–90 neurons at the output of the 4F system makes it possible to effectively solve classification problems, so a test set of handwritten numbers from the MNIST database is classified with up to 95% reliability. Practical significance. The results obtained give hope for the use of a 4F system used in the mode of a diffraction neural network to process video streams in real time.

Keywords:

4F scheme, diffraction neural network, convolutional neural network, object classification, spatial light modulator

Acknowledgements:

the research was carried out within the framework of the scientific program of the National Center for Physics and Mathematics, direction № 1 “National Center for Research on Supercomputer Architectures”. Stage 2023–2025

OCIS codes: 070.0070, 050.0050

References:

1. Vander Lugt A. Signal detection by spatial complex filtering // IEEE Trans. Inf. Theory. 1964. V. 10. P. 139–145.
2. Престон К. Когерентные оптические вычислительные машины: Пер. с англ. Москва: Мир, 1974. 399 с.
Preston K. Coherent optical computing machines: Translated from English by Mir, 1974.
3. Акаев А.А., Майоров С.А. Когерентные оптические вычислительные машины. Ленинград: Машиностроение, 1977. 327 с.
Akaev A.A., Mayorov S.A. Coherent optical computing machines. Leningrad: Mashinostroenie, 1977. 327 p.
4. Кейсесент Д., Колфилд Х. Дж., Томпсон Б.Дж. Оптическая обработка информации / Под ред. Д. Кейсесента. М.: Мир, 1980. 349 с.
Casesent D., Caulfield H.J., Thompson B.J. Optical information processing / Ed. by D. Casesent. Moscow: Mir, 1980. 349 p.
5. Mahalanobis A., Kumar B.V.K.V., Sims S.R.F. Distance-classifier correlation filters for multiclass target recognition // Applied Optics. 1996. V. 35. № 17. P. 3127–3133. https://doi.org/10.1364/AO.35.003127
6. Javidi B., Horner J.L. Optical pattern recognition for validation and security verification // Optical engineering. 1994. V. 33. № 6. P. 1752–1756. https://doi.org/10.1117/12.170736
7. Volodin B.L., Kippelen B., Meerholz K. et al. A polymeric optical pattern-recognition system for security verification // Nature. 1996. V. 383. № 6595. P. 58–60.
8. Qadri M.T., Asif M. Automatic number plate recognition system for vehicle identification using optical character recognition // 2009 international conference on education technology and computer. IEEE. 2009. P. 335–338. https://doi.org/10.1109/ICETC.2009.54
9. Savvides M., Kumar B.V.K.V., Khosla P.K. Cancelable biometric filters for face recognition // Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004. IEEE. 2004. V. 3. P. 922–925. https://doi.org/10.1109/ICPR.2004.1334679
10. Schönleber M. Joint transform correlator subtracting a modified Fourier spectrum / M. Schönleber, G. Cedilnik, H.-J. Tiziani // Applied Optics. 1995. V. 34. № 32. P. 7532–7537. https://doi.org/10.1364/AO.34.007532
11. Mahalanobis A., Kumar B.V.K.V., Song S., Sims S.R.F., Epperson J.F. Unconstrained correlation filters // Applied Optics. 1994. V. 33. № 17. P. 3751–3759. https://doi.org/10.1364/AO.33.003751
12. Refregier P. Optimal trade-off filters for noise robustness, sharpness of the correlation peak, and Horner efficiency // Optics Letters. 1991. V. 16. № 11. P. 829–831. https://doi.org/10.1364/OL.16.000829
13. Evtikhiev N.N., Starikov N.S., Shaulskiy D.V., Starikov R.S., Zlokazov E.Yu. Invariant correlation filter with linear phase coefficient holographic realization in 4F correlator // Optical Engineering. 2011. V. 50. № 6. P. 065803-065803-5. https://doi.org/10.1117/1.3592518
14. Ge P., Li Q., Feng H., Xu Z. Image rotation and translation measurement based on double phase-encoded joint transform correlator // Applied Optics. 2011. V. 50. №27. P. 5235–5242. https://doi.org/10.1016/j.ijleo.2015.02.012
15. Lin X., Rivenson Y., Yardimci N.T., Veli M., Luo Y., Jarrahi M., Ozcan A. All-optical machine learning using diffractive deep neural networks // Science. 2018. V. 361. P. 1004–1008. https://doi.org/10.1126/science.aat8084
16. Denz C. Optical neural networks. Springer Science & Business Media, Vieweg Verlag. 1998. 458 p.
17. Lin X., Rivenson Y., Yardimci N.T., Veli M., Luo Y., Jarrahi M., Ozcan A. All-optical machine learning using diffractive deep neural networks // Science. 2018. V. 361. P. 1004–1008. https://doi.org/10.1126/science.aat8084
18. Shen Y., Harris N., Skirlo S., Prabhu M., Baehr-Jones T., Hochberg M., Sun X., Zhao S., Larochelle H., Englund D., Soljačić M. Deep learning with coherent nanophotonic circuits // Nature Photonics. 2017. V. 11. P. 441–446. https://doi.org/ 10.1038/nphoton.2017.93
19. Zhou T., Fang L., Yan T., Wu J., Li Y., Fan J., Wu H., Lin X., Dai Q. In situ optical backpropagation training of diffractive optical neural networks // Photonics Research. 2020. V. 8. № 6. P. 940–953. https://doi.org/10.1364/PRJ.389553
20. Chang J., Sitzmann V., Dun X., Heidrich W., Wetzstein G. Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification // Scientific Reports. 2018. V. 8. 12324. https://doi.org/ 10.1038/s41598-018-30619-y
21. Doskolovich L. L., Soshnikov D.V., Motz G.A., Byzov E.V., Bezus E.A., Bykov D.A., Kazanskiy N.L. Design of cascaded does for focusing different wavelengths to different points // Photonics. 2024. V. 11. № 9. P. 791. https://doi.org/10.3390/photonics11090791
22. Сошников Д.В., Досколович Л.Л., Бызов Е.В. Градиентный метод расчета каскадных ДОЭ и его применение в задаче классификации рукописных цифр // Компьютерная оптика. 2023. Т. 47. № 5. С. 691–701. https://doi.org/10.18287/2412-6179-CO-1314
Soshnikov D.V., Doskolovich L.L., Byzov E.V. Gradient method of cascade DOE calculation and its application in the problem of classification of handwritten digits // Computer Optics. 2023. V. 47. № 5. P. 691–701.
https://doi.org/10.18287/2412-6179-CO-1314
23. Электронный ресурс URL: https://technicalcity. b-cdn.net/ru/video/Tesla-V100-PCIe-32-GB-protivH100-PCIe-80-GB (дата обращения 04.03.2025)
Electronic resource URL: https://technicalcity.b-cdn.net/ru/video/Tesla-V100-PCIe-32-GB-protiv-H100-PCIe-80-GB (accessed 04.03.2025)