Please use this identifier to cite or link to this item:
|Title:||HolyLight : a nanophotonic accelerator for deep learning in data centers||Authors:||Liu, Weichen
|Keywords:||Engineering::Computer science and engineering||Issue Date:||2019||Source:||Liu, W., Liu, W., Ye, Y., Lou, Q., Xie, Y., & Jiang, L. (2019). HolyLight : a nanophotonic accelerator for deep learning in data centers. Proceedings of the Design, Automation & Test in Europe Conference & Exhibition (DATE), 1483-1488. doi:10.23919/DATE.2019.8715195||Project:||04INS000515C130OST01
|Abstract:||Convolutional Neural Networks (CNNs) are widely adopted in object recognition, speech processing and machine translation, due to their extremely high inference accuracy. However, it is challenging to compute massive computationally expensive convolutions of deep CNNs on traditional CPUs and GPUs. Emerging Nanophotonic technology has been employed for on-chip data communication, because of its CMOS compatibility, high bandwidth and low power consumption. In this paper, we propose a nanophotonic accelerator, HolyLight, to boost the CNN inference throughput in datacenters. Instead of an all-photonic design, HolyLight performs convolutions by photonic integrated circuits, and process the other operations in CNNs by CMOS circuits for high inference accuracy. We first build HolyLight-M by microdisk-based matrix-vector multipliers. We find analog-to-digital converters (ADCs) seriously limit its inference throughput per Watt. We further use microdisk-based adders and shifters to architect HolyLight-A without ADCs. Compared to the state-of-the-art ReRAM-based accelerator, HolyLight-A improves the CNN inference throughput per Watt by 13× with trivial accuracy degradation.||URI:||https://hdl.handle.net/10356/145393||DOI:||10.23919/DATE.2019.8715195||Rights:||© 2019 European Design and Automation Association (EDAA). All rights reserved. This paper was published in 2019 Design, Automation & Test in Europe Conference & Exhibition (DATE) and is made available with permission of European Design and Automation Association (EDAA).||Fulltext Permission:||open||Fulltext Availability:||With Fulltext|
|Appears in Collections:||SCSE Conference Papers|
Updated on Apr 19, 2021
Updated on Apr 19, 2021
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.