Saugata Ghose
Assistant Professor – Department of Computer Science – University of Illinois Urbana-Champaign
Invited Talk: Using Processing-in-Memory to Accelerate Edge Machine Learning
Slides
Video
Abstract:
Machine learning (ML) is often thought of as a computationally intensive workload, and there has been much work focusing on how to accelerate the core computation of ML algorithms. However, as ML models grow in complexity, their model footprints exceed the limited storage capacity of many accelerators. In this talk, we will start by exploring the performance of and bottlenecks associated with edge neural network (NN) models, discussing the results of a detailed characterization that we conducted. Then, we will discuss how the memory system plays an outsized role in these bottlenecks, and will explore two types of processing-in-memory (PIM) architectures (Mensa and RACER) that can alleviate these bottlenecks. As we will show, both Mensa and RACER offer alternatives to conventional monolithic ML accelerators, and in doing so significantly improve the performance and energy consumption of edge NN model inference.
Biography:
Saugata Ghose is an assistant professor in the Department of Computer Science at the University of Illinois Urbana-Champaign, where he leads the ARCANA Research Group. He holds M.S. and Ph.D. degrees in electrical and computer engineering from Cornell University, and dual B.S. degrees in computer science and in computer engineering from Binghamton University, State University of New York. Prior to joining Illinois, he was a postdoc and later a systems scientist at Carnegie Mellon University. He received the best paper award from DFRWS-EU in 2017 for work on solid-state drive forensics, and was a 2019 Wimmer Faculty Fellow at CMU. His current research interests include data-oriented computer architectures and systems, new interfaces between systems software and hardware, energy-efficient memory and storage, and architectures for emerging platforms and domains. For more information, please visit his website at https://ghose.cs.illinois.edu.