eduzhai > Applied Sciences > Engineering >

Massively Parallel Amplitude-Only Fourier Neural Network

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 28 pages

Abstract: Machine-intelligence has become a driving factor in modern society. However,its demand outpaces the underlying electronic technology due to limitationsgiven by fundamental physics such as capacitive charging of wires, but also bysystem architecture of storing and handling data, both driving recent trendstowards processor heterogeneity. Here we introduce a novel amplitude-onlyFourier-optical processor paradigm capable of processing large-scale ~(1,000 x1,000) matrices in a single time-step and 100 microsecond-short latency.Conceptually, the information-flow direction is orthogonal to thetwo-dimensional programmable-network, which leverages 10^6-parallel channels ofdisplay technology, and enables a prototype demonstration performingconvolutions as pixel-wise multiplications in the Fourier domain reaching petaoperations per second throughputs. The required real-to-Fourier domaintransformations are performed passively by optical lenses at zero-static power.We exemplary realize a convolutional neural network (CNN) performingclassification tasks on 2-Megapixel large matrices at 10 kHz rates, whichlatency-outperforms current GPU and phase-based display technology by one andtwo orders of magnitude, respectively. Training this optical convolutionallayer on image classification tasks and utilizing it in a hybridoptical-electronic CNN, shows classification accuracy of 98 (MNIST) and 54 (CIFAR-10). Interestingly, the amplitude-only CNN is inherently robust againstcoherence noise in contrast to phase-based paradigms and features an over 2orders of magnitude lower delay than liquid crystal-based systems. Beyondcontributing to novel accelerator technology, scientifically thisamplitude-only massively-parallel optical compute-paradigm can be far-reachingas it de-validates the assumption that phase-information outweighs amplitude inoptical processors for machine-intelligence.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×