講題: Accelerating AI Applications based on High-Performance, High-Accuracy Approximate Computing

時間日期地點:  2/27 (Wed) 3:30-5:00PM in ED 117

講員名字&簡歷:

Kai-Chiang Wu received the B.S. and M.S. degrees in computer science from National Tsing Hua University, Hsinchu, Taiwan, in 2002 and 2004, respectively, and the Ph.D. degree in electrical and computer engineering from Carnegie Mellon University, Pittsburgh, PA, in 2011. From 2011 to 2013, he was with Intel Corporation, Hillsboro, OR, as a senior software engineer working on the research and development of CAD tools for reliability verification. He is currently an assistant professor in the Department of Computer Science, National Chiao Tung University, Hsinchu, Taiwan. His research interests include AI computing platform design, approximate computing for AI applications, reliable/trustworthy system design, and algorithms for electronic design automation. He was the recipient of a Best Paper Award from IEEE International Conference on Computer Design (ICCD 2008).

大綱:

Approximate computing (AC) is an emerging strategy which trades computational accuracy for computational cost in terms of performance, energy, and/or area. In this talk, I will give an introductory overview of AC and illustrate how AC can be applied for high-performance energy-efficient arithmetic computation, while considering the accuracy requirement of error-tolerant (e.g., deep learning) applications. Then, I will present novel AC arithmetic component and architecture, which are the first research studies using in-situ sensors for accelerating AI applications based on AC, and outperform the prior art. Finally, some interesting and promising experiments and results will be demonstrated, and a life demo will be given.