Softmax regression design for stochastic computing based deep convolutional neural networks

Zihao Yuan, Ji Li, Zhe Li, Caiwen Ding, Ao Ren, Bo Yuan, Qinru Qiu, Jeffrey Draper, Yanzhi Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

9 Scopus citations

Abstract

Recently, Deep Convolutional Neural Networks (DCNNs) have made tremendous advances, achieving close to or even better accuracy than human-level perception in various tasks. Stochastic Computing (SC), as an alternate to the conventional binary computing paradigm, has the potential to enable massively parallel and highly scalable hardware implementations of DCNNs. In this paper, we design and optimize the SC based Softmax Regression function. Experiment results show that compared with a binary SR, the proposed SC-SR under longer bit stream can reach the same level of accuracy with the improvement of 295X, 62X, 2617X in terms of power, area and energy, respectively. Binary SR is suggested for future DCNNs with short bit stream length input whereas SC-SR is recommended for longer bit stream.

Original languageEnglish (US)
Title of host publicationGLSVLSI 2017 - Proceedings of the Great Lakes Symposium on VLSI 2017
PublisherAssociation for Computing Machinery
Pages467-470
Number of pages4
VolumePart F127756
ISBN (Electronic)9781450349727
DOIs
StatePublished - May 10 2017
Event27th Great Lakes Symposium on VLSI, GLSVLSI 2017 - Banff, Canada
Duration: May 10 2017May 12 2017

Other

Other27th Great Lakes Symposium on VLSI, GLSVLSI 2017
CountryCanada
CityBanff
Period5/10/175/12/17

    Fingerprint

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Yuan, Z., Li, J., Li, Z., Ding, C., Ren, A., Yuan, B., Qiu, Q., Draper, J., & Wang, Y. (2017). Softmax regression design for stochastic computing based deep convolutional neural networks. In GLSVLSI 2017 - Proceedings of the Great Lakes Symposium on VLSI 2017 (Vol. Part F127756, pp. 467-470). Association for Computing Machinery. https://doi.org/10.1145/3060403.3060467