Dissertation Defense: Accuracy Analysis and Design Optimization for Low-Cost High-Accuracy Stochastic Circuit

Date: 2023/05/23 - 2023/05/23

Dissertation Title: Accuracy Analysis and Design Optimization for Low-Cost High-Accuracy Stochastic Circuit

Speaker: Kuncai Zhong, Ph.D. candidate at UM-SJTU Joint Institute

Time: May 23rd from 2:00 p.m., 2023 (Beijing Time)

Location: Room 503, Longbin Building

Abstract

Stochastic computing (SC) is an unconventional computing paradigm and does computation on values encoded by stochastic bit streams. Based on the special encoding way, SC can generally lead to low-cost computation circuitry and high fault tolerance. It has emerged as a promising solution to design digital circuits in the post-Moore era, where an SC circuit typically consists of a randomizer to generate stochastic bit streams and an SC core to do computation based on them. However, recent SC circuit designs either have a low accuracy or high hardware cost. They usually need a complex randomizer to ensure a high accuracy, which increases the overall cost of the circuit and occupies most area. Furthermore, they are also generally sub-optimal due to the separate optimization of the randomizer and the SC core. To address these issues, this dissertation studies the analysis and design of low-cost high-accuracy SC circuits.

Biography

Kuncai Zhong received his B.S. degree from Xi'an Jiaotong University in 2018. He has been working towards a Ph.D. degree since 2018 under supervision of Prof. Weikang Qian, Shanghai Jiao Tong University. He is interested in stochastic computing.