
Proceedings Paper
Electro-optic perceptron towards 10^18 MAC/J-efficient photonic neural networks (Conference Presentation)
Paper Abstract
Non-van Neumann compute engines such as neuromorphic electronics have shown to outperform CPUs by 3-4 orders of magnitude in terms of ‘weighted addition’, namely multiply-accumulate (MAC)-per-Joule. Here, we discuss experimental devices for a photonic neural network (NN) with an energy efficiency targeting10^18 MAC/J. We consider an electro-optic perceptron consisting of a photodetector (summation) coupled to an EO modulator (nonlinear activation function, NLAF) [George et al, Opt.Exp. 2019]. The perceptron’s efficiency is proportional to the electronic charge at the NLAF; in case of Silicon MZI modulators, this is ~10^6 charges hence the MAC/J is similar to TrueNorth. However, co-integration of emerging EO materials such as ITO into Si MZIs enables efficient modulation (e.g. VpL=0.5 V-mm [Armin et al, APL Phot. 2018]. Here we discuss latest results of a ITO-Silicon MZM with a record-low VpL=0.06 V-mm, and show noise-based NN training results of our in-house software PhotonFlow.
Paper Details
Date Published: 9 March 2020
PDF
Proc. SPIE 11299, AI and Optical Data Sciences, 112990D (9 March 2020); doi: 10.1117/12.2546966
Published in SPIE Proceedings Vol. 11299:
AI and Optical Data Sciences
Bahram Jalali; Ken-ichi Kitayama, Editor(s)
Proc. SPIE 11299, AI and Optical Data Sciences, 112990D (9 March 2020); doi: 10.1117/12.2546966
Show Author Affiliations
Rubab Amin, The George Washington Univ. (United States)
Mario Miscuglio, The George Washington Univ. (United States)
Bhavin J. Shastri, Princeton Univ. (United States)
Mario Miscuglio, The George Washington Univ. (United States)
Bhavin J. Shastri, Princeton Univ. (United States)
Paul Prucnal, Princeton Univ. (United States)
Volker J. Sorger, The George Washington Univ. (United States)
Volker J. Sorger, The George Washington Univ. (United States)
Published in SPIE Proceedings Vol. 11299:
AI and Optical Data Sciences
Bahram Jalali; Ken-ichi Kitayama, Editor(s)
© SPIE. Terms of Use
