Talks & Seminars
Title: Managing Data Movement in Neural Network Accelerators
Prof. Rajeev Balasubramonian, University of Utah
Date & Time: January 3, 2019 17:00
Venue: Department of Computer Science and Engineering, F. C. Kohli Auditorium, 01st Floor, 'B' Block, Kanwal Rekhi (KReSIT) Building
Industry has invested heavily in accelerators for machine learning, and deep neural networks in particular. Google TPUs are used heavily for inference and training in Google datacenters, modern phones already incorporate several accelerators for a variety of tasks, NVIDIA's Volta GPU has hundreds of tensor cores, Intel's Mobileye and Alphabet's Waymo are developing technologies for self-driving cars, and Intel's Movidius Neural Stick brings efficient inference to edge devices. The talk will first provide an overview of innovations in this accelerator landscape. It will then describe a few approaches to alleviate the primary bottleneck in such systems, namely, the high cost of data movement. In particular, emerging resistive memories can be leveraged to perform in-situ dot-product operations, and new architectures/dataflows can be designed to reduce the distances traveled by operands.
Speaker Profile:
Rajeev Balasubramonian is a Professor at the School of Computing, University of Utah. He received his B.Tech in Computer Science and Engineering from the Indian Institute of Technology, Bombay in 1998. He received his MS (2000) and Ph.D. (2003) degrees from the University of Rochester. His primary research interests include memory systems, security, and application-specific architectures, and his work appears regularly at the top architecture conferences. Prof. Balasubramonian is a recipient of a US National Science Foundation CAREER award, an IBM Faculty Partnership award, an HP Innovation Research Program award, an Intel Outstanding Research Award, various teaching awards at the University of Utah, and multiple best paper awards.
List of Talks


Faculty CSE IT
Forgot Password
    [+] Sitemap     Feedback