Neural Schr\"{o}dinger Equation:Physical Law as Neural Network

We show a new family of neural networks based on the Schrodinger equation (SE-NET). In this analogy, the trainable weights of the neural networks correspond to the physical quantities of the Schrodinger equation. These physical quantities can be trained using the complex-valued adjoint method. Since the propagation of the SE-NET can be described by the evolution of physical systems, its outputs can be computed by using a physical solver. As a demonstration, we implemented the SE-NET using the finite difference method. The trained network is transferable to actual optical systems. Based on this concept, we show a numerical demonstration of end-to-end machine learning with an optical frontend. Our results extend the application field of machine learning to hybrid physical-digital optimizations.

[1]  H Takahashi,et al.  Optical circuit design based on a wavefront-matching method. , 2005, Optics letters.

[2]  Sandeep Subramanian,et al.  Deep Complex Networks , 2017, ICLR.

[3]  Adam S. Backer.  Computational inverse design for cascaded systems of metasurface optics. , 2019, Optics express.

[4]  Paul R. Prucnal,et al.  Demonstration of Photonic Neural Network for Fiber Nonlinearity Compensation in Long-Haul Transmission Systems , 2020, 2020 Optical Fiber Communications Conference and Exhibition (OFC).

[5]  David Duvenaud,et al.  Neural Ordinary Differential Equations , 2018, NeurIPS.

[6]  Yann Le Cun,et al.  A Theoretical Framework for Back-Propagation , 1988 .

[7]  Markus Heinonen,et al.  ODE$^2$VAE: Deep generative second order ODEs with Bayesian neural networks , 2019, NeurIPS.

[8]  Yi Luo,et al.  All-optical machine learning using diffractive deep neural networks , 2018, Science.

[9]  Yann LeCun,et al.  The mnist database of handwritten digits , 2005 .

[10]  Yuval Tassa,et al.  Continuous control with deep reinforcement learning , 2015, ICLR.

[11]  T. Hashimoto,et al.  New Optical Waveguide Design Based on Wavefront Matching Method , 2007, Journal of Lightwave Technology.

[12]  Qianshun Chang,et al.  Difference Schemes for Solving the Generalized Nonlinear Schrödinger Equation , 1999 .

[13]  Qionghai Dai,et al.  Supplementary Document : Spatial-spectral Encoded Compressive Hyperspectral Imaging , 2014 .

[14]  Eldad Haber,et al.  Deep Neural Networks Motivated by Partial Differential Equations , 2018, Journal of Mathematical Imaging and Vision.

[15]  Stephen P. Boyd,et al.  End-to-end optimization of optics and image processing for achromatic extended depth of field and super-resolution imaging , 2018, ACM Trans. Graph..

[16]  Dirk Englund,et al.  Deep learning with coherent nanophotonic circuits , 2017, 2017 Fifth Berkeley Symposium on Energy Efficient Electronic Systems & Steep Transistors Workshop (E3S).

[17]  Qian Chen,et al.  A Review of Optical Neural Networks , 2020, IEEE Access.

[18]  T. Okino.  Correlation between Diffusion Equation and Schrödinger Equation , 2013 .

[19]  Markus Heinonen,et al.  ODE2VAE: Deep generative second order ODEs with Bayesian neural networks , 2019, NeurIPS.

[20]  Raja Giryes,et al.  Learned phase coded aperture for the benefit of depth of field extension. , 2018, Optics express.

[21]  Gordon Wetzstein,et al.  Deep Optics for Monocular Depth Estimation and 3D Object Detection , 2019, 2019 IEEE/CVF International Conference on Computer Vision (ICCV).

[22]  Demis Hassabis,et al.  Mastering the game of Go with deep neural networks and tree search , 2016, Nature.

[23]  Adrian Stern,et al.  Compressive sensing spectrometry based on liquid crystal devices. , 2013, Optics letters.

[24]  Heung-No Lee,et al.  Convolutional neural networks for the reconstruction of spectra in compressive sensing spectrometers , 2019, OPTO.

[25]  Kristan Temme,et al.  Supervised learning with quantum-enhanced feature spaces , 2018, Nature.

[26]  H. Haus.  Waves and fields in optoelectronics , 1983 .

[27]  Vincent Y. F. Tan,et al.  On Robustness of Neural Ordinary Differential Equations , 2020, ICLR.

[28]  Shanhui Fan,et al.  Adjoint Method and Inverse Design for Nonlinear Nanophotonic Devices , 2018, ACS Photonics.

[29]  Ginés Lifante Pedrola.  Beam Propagation Method for Design of Optical Waveguide Devices: Pedrola/Beam Propagation Method for Design of Optical Waveguide Devices , 2015 .

[30]  Eldad Haber,et al.  Reversible Architectures for Arbitrarily Deep Residual Neural Networks , 2017, AAAI.

[31]  Shanhui Fan,et al.  Training of Photonic Neural Networks through In Situ Backpropagation , 2018, 2019 Conference on Lasers and Electro-Optics (CLEO).

[32]  Serge J. Belongie,et al.  Residual Networks Behave Like Ensembles of Relatively Shallow Networks , 2016, NIPS.

[33]  Stefano Soatto,et al.  Deep relaxation: partial differential equations for optimizing deep neural networks , 2017, Research in the Mathematical Sciences.

[34]  Ian A. D. Williamson,et al.  Training of Photonic Neural Networks through In Situ Backpropagation , 2019 .

[35]  Keisuke Fujii,et al.  Quantum circuit learning , 2018, Physical Review A.

[36]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[37]  Yoshua Bengio,et al.  Unitary Evolution Recurrent Neural Networks , 2015, ICML.

[38]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[39]  J. S. Aitchison,et al.  Discrete Spatial Optical Solitons in Waveguide Arrays , 1998 .

[40]  Gordon Wetzstein,et al.  Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification , 2018, Scientific Reports.

[41]  Heung-No Lee,et al.  Compressive Sensing Spectroscopy Using a Residual Convolutional Neural Network , 2020, Sensors.

[42]  David Duvenaud,et al.  FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models , 2018, ICLR.

[43]  Michiel Hermans,et al.  Embodiment of Learning in Electro-Optical Signal Processors , 2016, Physical review letters.

[44]  Adrian Stern,et al.  Compressive hyperspectral image reconstruction with deep neural networks , 2019 .