Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
164 tokens/sec
GPT-4o
10 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
5 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Low-power scalable multilayer optoelectronic neural networks enabled with incoherent light (2402.01988v1)

Published 3 Feb 2024 in cs.ET and physics.optics

Abstract: Optical approaches have made great strides towards the goal of high-speed, energy-efficient computing necessary for modern deep learning and AI applications. Read-in and read-out of data, however, limit the overall performance of existing approaches. This study introduces a multilayer optoelectronic computing framework that alternates between optical and optoelectronic layers to implement matrix-vector multiplications and rectified linear functions, respectively. Our framework is designed for real-time, parallelized operations, leveraging 2D arrays of LEDs and photodetectors connected via independent analog electronics. We experimentally demonstrate this approach using a system with a three-layer network with two hidden layers and operate it to recognize images from the MNIST database with a recognition accuracy of 92% and classify classes from a nonlinear spiral data with 86% accuracy. By implementing multiple layers of a deep neural network simultaneously, our approach significantly reduces the number of read-ins and read-outs required and paves the way for scalable optical accelerators requiring ultra low energy.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (46)
  1. Deep learning. Nature 521, 436–444 (2015).
  2. AI and compute, 2018. https://openai.com/blog/ai-and-compute 4 (2018).
  3. Sevilla, J. et al. Compute trends across three eras of machine learning. arXiv preprint arXiv:2202.05924 (2022).
  4. Compute and energy consumption trends in deep learning inference. arXiv preprint arXiv:2109.05472 (2021).
  5. Horowitz, M. 1.1 computing’s energy problem (and what we can do about it). In 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), 10–14 (IEEE, 2014).
  6. Christensen, D. V. et al. 2022 roadmap on neuromorphic computing and engineering. Neuromorphic Computing and Engineering (2022).
  7. Why future supercomputing requires optics. Nature Photonics 4, 261–263 (2010).
  8. Fully parallel, high-speed incoherent optical method for performing discrete fourier transforms. Optics Letters 2, 1–3 (1978).
  9. Optical implementation of the hopfield model. Applied Optics 24, 1469–1475 (1985).
  10. Optical neurochip based on a three-layered feed-forward model. Optics Letters 15, 1362–1364 (1990).
  11. Adaptive optical networks using photorefractive crystals. Applied Optics 27, 1752–1759 (1988).
  12. Denz, C. Optical neural networks (Springer Science & Business Media, 2013).
  13. Miller, D. A. Attojoule optoelectronics for low-energy information processing and communications. Journal of Lightwave Technology 35, 346–396 (2017).
  14. Wetzstein, G. et al. Inference in artificial intelligence with deep optics and photonics. Nature 588, 39–47 (2020).
  15. Shastri, B. J. et al. Photonics for artificial intelligence and neuromorphic computing. Nature Photonics 15, 102–114 (2021).
  16. Large-scale optical neural networks based on photoelectric multiplication. Physical Review X 9, 021032 (2019).
  17. Bogaerts, W. et al. Programmable photonic circuits. Nature 586, 207–216 (2020).
  18. Tait, A. N. et al. Neuromorphic photonic networks using silicon photonic weight banks. Scientific Reports 7, 1–10 (2017).
  19. Tait, A. N. et al. Silicon photonic modulator neuron. Physical Review Applied 11, 064043 (2019).
  20. Nahmias, M. A. et al. Photonic multiply-accumulate operations for neural networks. IEEE Journal of Selected Topics in Quantum Electronics 26, 1–18 (2019).
  21. Feldmann, J. et al. Parallel convolutional processing using an integrated photonic tensor core. Nature 589, 52–58 (2021).
  22. Xu, X. et al. 11 tops photonic convolutional accelerator for optical neural networks. Nature 589, 44–51 (2021).
  23. An on-chip photonic deep neural network for image classification. Nature 1–6 (2022).
  24. Lin, X. et al. All-optical machine learning using diffractive deep neural networks. Science 361, 1004–1008 (2018).
  25. Miscuglio, M. et al. Massively parallel amplitude-only fourier neural network. Optica 7, 1812–1819 (2020).
  26. Fully reconfigurable coherent optical vector–matrix multiplication. Optics Letters 45, 5752–5755 (2020).
  27. Bernstein, L. et al. Freely scalable and reconfigurable optical hardware for deep learning. Scientific reports 11, 1–12 (2021).
  28. Zhou, T. et al. Large-scale neuromorphic optoelectronic computing with a reconfigurable diffractive processing unit. Nature Photonics 15, 367–373 (2021).
  29. Bernstein, L. et al. Single-shot optical neural network. arXiv preprint arXiv:2205.09103 (2022).
  30. Wang, T. et al. An optical neural network using less than 1 photon per multiplication. Nature Communications 13, 1–8 (2022).
  31. Zhang, H. et al. An optical neural chip for implementing complex-valued neural network. Nature Communications 12, 1–11 (2021).
  32. Zuo, Y. et al. All-optical neural network with nonlinear activation functions. Optica 6, 1132–1137 (2019).
  33. Li, G. H. et al. All-optical ultrafast relu function for energy-efficient nanophotonic deep learning. arXiv preprint arXiv:2201.03787 (2022).
  34. Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification. Scientific reports 8, 1–10 (2018).
  35. Wang, T. et al. Image sensing with multilayer nonlinear optical neural networks. Nature Photonics 17, 408–415 (2023).
  36. Williamson, I. A. et al. Reprogrammable electro-optic nonlinear activation functions for optical neural networks. IEEE Journal of Selected Topics in Quantum Electronics 26, 1–12 (2019).
  37. Photonic extreme learning machine by free-space optical propagation. Photonics Research 9, 1446–1454 (2021).
  38. Shi, W. et al. Loen: Lensless opto-electronic neural network empowered machine vision. Light: Science & Applications 11, 1–12 (2022).
  39. Miller, D. A. All linear optical devices are mode converters. Optics Express 20, 23985–23993 (2012).
  40. Optical complex media as universal reconfigurable linear operators. Optica 6, 465–472 (2019).
  41. All-optical synthesis of an arbitrary linear transformation using diffractive surfaces. Light: Science & Applications 10, 1–21 (2021).
  42. Arbitrary linear transformations for photons in the frequency synthetic dimension. Nature Communications 12, 1–9 (2021).
  43. Patterson, D. Good news about the carbon footprint of machine learning training. https://blog.research.google/2022/02/good-news-about-carbon-footprint-of.html (2022). URL https://blog.research.google/2022/02/good-news-about-carbon-footprint-of.html.
  44. Viswanathula, R. Global AI and data science. https://community.ibm.com/community/user/ai-datascience/blogs/rachana-vishwanathula/2023/05/04/estimating-chatgpts-carbon-footprint (2023).
  45. Mini-led, micro-led and oled displays: Present status and future perspectives. Light: Science & Applications 9, 1–16 (2020).
  46. Deng, L. The mnist database of handwritten digit images for machine learning research [best of the web]. IEEE Signal Processing Magazine 29, 141–142 (2012).

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets