一种无需非线性结构的光学卷积神经网络
Optical Convolutional Neural Network for Image Recognition without Nonlinear Structures
DOI: 10.12677/mos.2024.133351, PDF,    科研立项经费支持
作者: 江 奔, 张 薇, 许 涛：上海理工大学光电信息与计算机工程学院，上海
关键词: 光学卷积神经网络；激活函数；马赫–曾德尔干涉仪；Optical Convolutional Neural Networks； Activation Function； Mach-Zehnder Interferometer
摘要: 卷积神经网络在视觉处理方面具有独特的优势。最近的一些研究使用光学的卷积神经网络来实现更为快速和低功耗的图像处理系统。我们的工作提出了一种无需额外非线性结构的光学卷积神经网络，使用绝对值函数作为激活函数，设置了特定的三层卷积神经网络，该网络对MNIST和Fashion-MNIST数据集的识别准确率与目前普遍使用的激活函数相差不大。通过对软件仿真与硬件仿真的结果进行对比，发现图像经过第一层卷积和非线性后的结果误差不超过2%，输出结果验证了我们的光学卷积神经网络对图像处理的有效性。这为实现高效可编程的光学卷积神经网络提供了可行方案。
Abstract: Convolutional neural networks (CNNs) have unique advantages in visual processing. Some recent studies have used optical CNNs to realize faster and low-power image processing systems. This study proposes an optical CNN without the need for additional nonlinear structures, using the absolute value function as the activation function. A specific three-layer CNN developed in this study provided recognition accuracies for the MNIST and Fashion-MNIST datasets similar to those of the commonly used activation functions currently in use. By comparing the results of the software simulation with the hardware simulation, it is found that the error of its results after the first layer of convolution and nonlinearity was not more than 2%, and the output results verify the effectiveness of our optical CNN for image processing. This provides a viable solution for realizing efficient and programmable optical CNNs.
文章引用：江奔, 张薇, 许涛. 一种无需非线性结构的光学卷积神经网络[J]. 建模与仿真, 2024, 13(3): 3851-3860. https://doi.org/10.12677/mos.2024.133351

参考文献

[1] Malik, A.S., Boyko, O., Aktar, N. and Young, W.F. (2001) A Comparative Study of MR Imaging Profile of Titanium Pedicle Screws. Acta Radiologica, 42, 291-293.
[2] Sui, X., Wu, Q., Liu, J., Chen, Q. and Gu, G. (2020) A Review of Optical Neural Networks. IEEE Access, 8, 70773-70783. [Google Scholar] [CrossRef] 
[3] Bagherian, H., Skirlo, S., Shen, Y., Meng, H., Ceperic, V. and Soljacic, M. (2018) On-Chip Optical Convolutional Neural Networks. arXiv preprint arXiv: 1808.03303.
[4] Xu, X., Zhu, L., Zhuang, W., Lu, L. and Yuan, P. (2022) A Convolution Neural Network Implemented by Three 3× 3 Photonic Integrated Reconfigurable Linear Processors. Photonics, 9, 80. [Google Scholar] [CrossRef] 
[5] Xu, R., Lv, P., Xu, F. and Shi, Y. (2021) A Survey of Approaches for Implementing Optical Neural Networks. Optics & Laser Technology, 136, Article ID: 106787. [Google Scholar] [CrossRef] 
[6] Selden, A. (1967) Pulse Transmission through a Saturable Absorber. British Journal of Applied Physics, 18, Article 743. [Google Scholar] [CrossRef] 
[7] Soljačić, M., Ibanescu, M., Johnson, S.G., Fink, Y. and Joannopoulos, J.D. (2002) Optimal Bistable Switching in Nonlinear Photonic Crystals. Physical Review E, 66, Article ID: 055601. [Google Scholar] [CrossRef] 
[8] Cheng, Z., Tsang, H.K., Wang, X., Xu, K. and Xu, J.B. (2013) In-Plane Optical Absorption and Free Carrier Absorption in Graphene-on-Silicon Waveguides. IEEE Journal of Selected Topics in Quantum Electronics, 20, 43-48. [Google Scholar] [CrossRef] 
[9] Williamson, I.A., Hughes, T.W., Minkov, M., Bartlett, B., Pai, S. and Fan, S. (2019) Reprogrammable Electro-Optic Nonlinear Activation Functions for Optical Neural Networks. IEEE Journal of Selected Topics in Quantum Electronics, 26, 1-12. [Google Scholar] [CrossRef] 
[10] Feldmann, J., Youngblood, N., Wright, C.D., Bhaskaran, H. and Pernice, W.H. (2019) All-Optical Spiking Neurosynaptic Networks with Self-Learning Capabilities. Nature, 569, 208-214. [Google Scholar] [CrossRef] [PubMed]
[11] Bhatt, D., Patel, C., Talsania, H., Patel, J., Vaghela, R., Pandya, S., et al. (2021) CNN Variants for Computer Vision: History, Architecture, Application, Challenges and Future Scope. Electronics, 10, Article 2470. [Google Scholar] [CrossRef] 
[12] Springenberg, J.T., Dosovitskiy, A., Brox, T. and Riedmiller, M. (2014) Striving for Simplicity: The All Convolutional net. arXiv preprint arXiv: 1412.6806.
[13] Ruder, S. (2016) An Overview of Gradient Descent Optimization Algorithms. arXiv preprint arXiv: 1609.04747.
[14] Apicella, A., Donnarumma, F., Isgrò, F. and Prevete, R. (2021) A Survey on Modern Trainable Activation Functions. Neural Networks, 138, 14-32. [Google Scholar] [CrossRef] [PubMed]
[15] Connelly, M.J. (2007) Semiconductor Optical Amplifiers. Springer Science & Business, Berlin.
[16] Harris, N.C., Ma, Y., Mower, J., Baehr-Jones, T., Englund, D., Hochberg, M. and Galland, C. (2014) Efficient, Compact and Low Loss Thermo-Optic Phase Shifter in Silicon. Optics Express, 22, 10487-10493. [Google Scholar] [CrossRef] 
[17] Reck, M., Zeilinger, A., Bernstein, H.J. and Bertani, P. (1994) Experimental Realization of Any Discrete Unitary Operator. Physical Review Letters, 73, 58. [Google Scholar] [CrossRef] 
[18] Clements, W.R., Humphreys, P.C., Metcalf, B.J., Kolthammer, W.S. and Walmsley, I.A. (2016) Optimal Design for Universal Multiport Interferometers. Optica, 3, 1460-1465. [Google Scholar] [CrossRef] 
[19] Chrostowski, L. and Hochberg, M. (2015) Silicon Photonics Design: From Devices to Systems. Cambridge University Press, Cambridge. [Google Scholar] [CrossRef] 
[20] Shen, Y., Harris, N. C., Skirlo, S., Prabhu, M., Baehr-Jones, T., Hochberg, M., et al. (2017) Deep Learning with Coherent Nanophotonic Circuits. Nature Photonics, 11, 441-446. [Google Scholar] [CrossRef] 
[21] Chrostowski, L., Shoman, H., Hammood, M., Yun, H., Jhoja, J., Luan, E., et al. (2019) Silicon Photonic Circuit Design Using Rapid Prototyping Foundry Process Design Kits. IEEE Journal of Selected Topics in Quantum Electronics, 25, 1-26. [Google Scholar] [CrossRef] 

为你推荐