Machine Learning and Systems a Virtuous Cycle
Kunle Olukotun (Stanford University)
Distinguished Lecture Series
Thursday, December 3, 2020, 3:30 pm
Zoom Meeting
Abstract
This talk is about the virtuous interplay between machine learning (ML) and systems. I will show examples of how high-performance systems can be used to train more accurate ML models and how ML models can be used to improve upon the ad-hoc heuristics used in complex system design and management. I will begin with a discussion of Domain Specific Architectures (DSAs) for ML and the design of the Plasticine reconfigurable dataflow architecture. Plasticine is composed of configurable compute units that can exploit nested-pipelined parallelism which is prevalent in ML models and configurable memory system that captures data locality and sustains compute throughput with multiple banking modes. The Plasticine research has spawned a commercial ML accelerator with outstanding performance and flexibility. I will describe how the Plasticine architecture can be used to build Taurus, an intelligent network data plane that enables ML models to be used to manage computer networks at full line-rate bandwidths. Lastly, I will describe how Bayesian Optimization can be used to improve programs written for Plasticine and to provide a high-level interface to Taurus for network programmers.
Bio
Kunle Olukotun is the Cadence Design Professor of Electrical Engineering and Computer Science at Stanford University. Olukotun is well known as a pioneer in multicore processor design and the leader of the Stanford Hydra chip multipocessor (CMP) research project. Olukotun founded Afara Websystems to develop high-throughput, low-power multicore processors for server systems. The Afara multicore processor, called Niagara, was acquired by Sun Microsystems. Niagara derived processors now power all Oracle SPARC-based servers. Olukotun currently directs the Stanford Pervasive Parallelism Lab (PPL), which seeks to proliferate the use of heterogeneous parallelism in all application areas using Domain Specific Languages (DSLs). Olukotun is a member of the Data Analytics for What's Next (DAWN) Lab which is developing infrastructure for usable machine learning. Olukotun is a co-founder of SambaNova Systems. Olukotun is an ACM Fellow and IEEE Fellow for contributions to multiprocessors on a chip and multi-threaded processor design and is the recipient of of the 2018 IEEE Harry H. Goode Memorial Award. Olukotun received his Ph.D. in Computer Engineering from the University of Michigan.