mlsanityData to 14 Oct 2026

Experimentally realized in situ backpropagation for deep learning in nanophotonic neural networks

arXiv 2205.08501 ↗First posted 17 May 2022cs.ET cs.LG physics.opticscited by 44

Not on the map. Why?

Abstract (arXiv)

Neural networks are widely deployed models across many scientific disciplines and commercial endeavors ranging from edge computing and sensing to large-scale signal processing in data centers. The most efficient and well-entrenched method to train such networks is backpropagation, or reverse-mode automatic differentiation. To counter an exponentially increasing energy budget in the artificial intelligence sector, there has been recent interest in analog implementations of neural networks, specifically nanophotonic neural networks for which no analog backpropagation demonstration exists. We design mass-manufacturable silicon photonic neural networks that alternately cascade our custom designed "photonic mesh" accelerator with digitally implemented nonlinearities. These reconfigurable photonic meshes program computationally intensive arbitrary matrix multiplication by setting physical voltages that tune the interference of optically encoded input data propagating through integrated Mach-Zehnder interferometer networks. Here, using our packaged photonic chip, we demonstrate in situ backpropagation for the first time to solve classification tasks and evaluate a new protocol to keep the entire gradient measurement and update of physical device voltages in the analog domain, improving on past theoretical proposals. Our method is made possible by introducing three changes to typical photonic meshes: (1) measurements at optical "grating tap" monitors, (2) bidirectional optical signal propagation automated by fiber switch, and (3) universal generation and readout of optical amplitude and phase. After training, our classification achieves accuracies similar to digital equivalents even in presence of systematic error. Our findings suggest a new training paradigm for photonics-accelerated artificial intelligence based entirely on a physical analog of the popular backpropagation technique.

Key facts. Cited by 44 papers in the map (data through 14 Oct 2026). Builds on 8 of its 72 references that are in the map. First posted on arXiv 17 May 2022.

Builds on

8 of 72 references.

References · 8

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2016Direct Feedback Alignment Provides Learning in Deep Neural Networkscited by 240
  3. 2021Deep physical neural networks enabled by a backpropagation algorithm for arbitrary physical systemscited by 128
  4. 2018Training of photonic neural networks through in situ backpropagationcited by 45
  5. 2019Reprogrammable Electro-Optic Nonlinear Activation Functions for Optical Neural Networkscited by 19
  6. 2021Silicon Photonic Architecture for Training Deep Neural Networks with Direct Feedback Alignmentcited by 17
  7. 2022Hybrid training of optical neural networkscited by 15
  8. 2019Parallel fault-tolerant programming of an arbitrary feedforward photonic networkcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 7; 2024: 11; 2025: 13; 2026: 13.

44 papers.

Sort:

By area:

  1. 2024Training of Physical Neural Networkscited by 49
  2. 2023Backpropagation-free Training of Deep Physical Neural Networkscited by 42
  3. 2023Spatially Varying Nanophotonic Neural Networkscited by 12
  4. 2024Arbitrary control over multimode wave propagation for machine learningcited by 9
  5. 2024Quantum Equilibrium Propagation for efficient training of quantum systems based on Onsager reciprocitycited by 9
  6. 2023Training neural networks with end-to-end optical backpropagationcited by 7
  7. 2023M3ICRO: Machine Learning-Enabled Compact Photonic Tensor Core based on PRogrammable Multi-Operand Multimode Interferencecited by 5
  8. 2024Training all-mechanical neural networks for task learning through in situ backpropagationcited by 5
  9. 2024Poor Man's Training on MCUs: A Memory-Efficient Quantized Back-Propagation-Free Approachcited by 5
  10. 2023Real-Time FJ/MAC PDE Solvers via Tensorized, Back-Propagation-Free Optical PINN Trainingcited by 4
  11. 2023A Blueprint for Precise and Fault-Tolerant Analog Neural Networkscited by 3
  12. 2024Asymmetrical estimator for training encapsulated deep photonic neural networkscited by 3
32 more
  1. 2025Near-Equilibrium Propagation training in nonlinear wave systemscited by 3
  2. 2023Mirage: An RNS-Based Photonic Accelerator for DNN Trainingcited by 2
  3. 2024Multiplexed all-optical permutation operations using a reconfigurable diffractive optical networkcited by 2
  4. 2025Scalable Back-Propagation-Free Training of Optical Physics-Informed Neural Networkscited by 2
  5. 2025Model-free Optical Processors using In Situ Reinforcement Learning with Proximal Policy Optimizationcited by 2
  6. 2024DOCTOR: Dynamic On-Chip Temporal Variation Remediation Toward Self-Corrected Photonic Tensor Acceleratorscited by 1
  7. 2024Quantum-secure multiparty deep learningcited by 1
  8. 2025A ghost mechanism: An analytical model of abrupt learning in recurrent networkscited by 1
  9. 2025SP2RINT: Spatially-Decoupled Physics-Inspired Progressive Inverse Optimization for Scalable, PDE-Constrained Meta-Optical Neural Network Trainingcited by 1
  10. 2025Circuit realization and hardware linearization of monotone operator equilibrium networkscited by 1
  11. 2025Low-rank surrogate modeling and stochastic zero-order optimization for training of neural networks with black-box layerscited by 1
  12. 2026Training deep physical neural networks with local physical information bottleneckcited by 1
  13. 2026Unlocked Backpropagation using Wave Scatteringcited by 1
  14. 2026Local learning for stable backpropagation-free neural network training towards physical learningcited by 1
  15. 2024Thermal Crosstalk Modelling and Compensation Methods for Programmable Photonic Integrated Circuitscited by 0
  16. 2024Computational metaoptics for imagingcited by 0
  17. 2025Experimental Demonstration of an Optical Neural PDE Solver via On-Chip PINN Trainingcited by 0
  18. 2025Scaling of hardware-compatible perturbative training algorithmscited by 0
  19. 2025All-optical temporal integration mediated by subwavelength heat antennascited by 0
  20. 2025Programmable metasurfaces for future photonic artificial intelligencecited by 0
  21. 2025Machine Intelligence on Wireless Edge Networkscited by 0
  22. 2025Universality of physical neural networks with multivariate nonlinearitycited by 0
  23. 2026Laser interferometry as a robust neuromorphic platform for machine learningcited by 0
  24. 20263D aperture-engineered diffractive neural networks for super-resolution electromagnetic wave computingcited by 0
  25. 2026Hybridizing Equilibrium Propagation with Ising Machines for Efficient Energy-Based Learningcited by 0
  26. 2026Optimizing Energy-based Neural Network Training with Coherent Ising Machinecited by 0
  27. 2026LLM-Guided Neural Architecture Search for Robust Co-Design of Physical Neural Networkscited by 0
  28. 2026Optical Implementation of Equilibrium Propagation Using Spatial Photonic Ising Machinescited by 0
  29. 2026Inverse-designed meta processing units for multi-task near-field photonic computingcited by 0
  30. 2026Program-Synthesis-Driven Autodesign of Universal Unitary Operatorscited by 0
  31. 2026Constrained Co-Design for Photonic Bayesian Neural Networkscited by 0
  32. 2026Physical policy gradient theorem for in situ stochastic-adjoint trainingcited by 0