Fall 2021 Electrical Engineering and Computer Science (EECS) Seminar Series

"Heterogeneous Computing and Memory Systems for Tensor-Based Applications"

Jiawen Liu

UC Merced

Faculty Host: Prof. Dong Li

Abstract:
Tensors, which generalize matrices to more than two dimensions, are fundamental to many disciplines. Thus, improving the performance and scalability of tensor computations is essential in many fields and realistic applications. Emerging heterogeneous computing and memory systems are promising to deliver efficient and scalable tensor computations, due to their hardware heterogeneity.

This talk will demonstrate how to design heterogeneous computing and memory systems for tensor-based applications. First, I will introduce the first efficient and scalable heterogeneous computing system for tensor-based neural network training. Second, I will present the first efficient and scalable heterogeneous memory system for tensor-based scientific computing applications.

Speaker Bio:
Jiawen Liu is a Ph.D. candidate in Electrical Engineering and Computer Science at the University of California, Merced. His research interests lie in High Performance Computing, Computer Systems, and their intersection with Machine Learning. In particular, he focuses on designing efficient and scalable heterogeneous computing and memory systems in data centers for tensor-based Machine Learning and Scientific Computing. During his Ph.D., he has published 14 papers (with seven of them as the first author), out of which eight appeared at top venues, such as MICRO, PPoPP, ICS, ASPLOS, IPDPS, and EuroSys. He receives the ACM ASPLOS Distinguished Artifact Award in 2021. He had internships in multiple institutes, including the Oak Ridge National Laboratory, the Pacific Northwest National Laboratory, the Facebook Distributed AI, and Tencent America.  He has served as Artifact Evaluation Committees at top venues, such as SOSP, ASPLOS, and LCTES.