The Distinguished Lecture on “Robust In-eDRAM Computing for Deep Learning and Combinatorial Optimization” will take place as follows:
Date: 11 July 2024 (Thu)
Time: 10:30 a.m. – 11:30 a.m.
Venue: Research Building N21, GF, Room G013
The speaker is:
Prof. TANG Xiyuan, Assistant Professor, Peking University, Beijing, China
The Lecture is:
Robust In-eDRAM Computing for Deep Learning and Combinatorial Optimization
Abstract:
The rapid advancement in AI necessitates an escalating demand for computing resources. Mixed-signal in-memory computing emerges as a promising solution, offering substantial improvements in energy efficiency and throughput. This is primarily achieved through the reduction of repetitive memory accesses and the maximization of parallelism. However, SoTA analog CIM systems are often plagued by accuracy degradation, attributed to intrinsic variabilities. This talk introduces a new eDRAM cell, that unifies the programming and computing in current domain, thus eliminating various error sources in mixed-signal in-memory computing. It enables the accurate sub-uA multi-level cell design, which significantly boosts the computing energy efficiency while maintaining good accuracy.Two design examples will be discussed: 1) a robust in-eDRAM computing macro for deep learning that realizes the best 4b computing energy efficiency and 2) a continuous-time in-eDRAM Ising machine with 8-level coefficients and leaked feedback annealing technique.
Biography:
Prof. TANG Xiyuan received the B.Sc. degree (Hons.) from the School of Microelectronics, Shanghai Jiao Tong University, Shanghai, China, in 2012, and the M.S. and Ph.D. degree in electrical engineering from The University of Texas at Austin, Austin, TX, USA, in 2014 and 2019 respectively. He is currently an Assistant Professor at Peking University, Beijing, China. He was a Design Engineer with Silicon Laboratories, Austin, TX from 2014 to 2017, where he was involved in the RF receiver design. From 2019-2021, he was a postdoctoral researcher at the University of Texas at Austin, Austin, TX. He serves on the Technical Program Committees (TPC) of ISSCC. He also serves as an associate editor for IEEE Solid-State Circuits Letters. His research interests include data converters, low-power mixed-signal circuits, and edge data processing.
For more details, kindly find the event poster, abstract and bio.
