Large-Scale Optical Neural Networks Based on Photoelectric Multiplication
Name
PhysRevX.9.021032.pdf
Description
Published version
Size
835.52 KB
Format
Adobe PDF
Checksum (MD5)
1f01bb9118711a4f2a1b7189b719abab
Author(s)
Hamerly, Ryan
Bernstein, Liane
Sludds, Alexander
Soljačić, Marin
Englund, Dirk
Date Issued
2019
Journal
Physical Review X
Publisher
American Physical Society (APS)
Version
Final published version
Abstract
© 2019 authors. Published by the American Physical Society. Published by the American Physical Society under the terms of the »https://creativecommons.org/licenses/by/4.0/» Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI. Recent success in deep neural networks has generated strong interest in hardware accelerators to improve speed and energy consumption. This paper presents a new type of photonic accelerator based on coherent detection that is scalable to large (N106) networks and can be operated at high (gigahertz) speeds and very low (subattojoule) energies per multiply and accumulate (MAC), using the massive spatial multiplexing enabled by standard free-space optical components. In contrast to previous approaches, both weights and inputs are optically encoded so that the network can be reprogrammed and trained on the fly. Simulations of the network using models for digit and image classification reveal a "standard quantum limit" for optical neural networks, set by photodetector shot noise. This bound, which can be as low as 50 zJ/MAC, suggests that performance below the thermodynamic (Landauer) limit for digital irreversible computation is theoretically possible in this device. The proposed accelerator can implement both fully connected and convolutional networks. We also present a scheme for backpropagation and training that can be performed in the same hardware. This architecture will enable a new class of ultralow-energy processors for deep learning.
MIT Department
Massachusetts Institute of Technology. Research Laboratory of Electronics
Terms of Use
Creative Commons Attribution 4.0 International license
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1103/PHYSREVX.9.021032
