ECE Seminar Series – May 24 (Fri) @ 2:00pm: “Computing Beyond Boolean Logic Using Time, Stochasticity, and Geometry,” Matthew W. Daniels, Project Leader, NIST

Come at 1:00p for Cookies, Coffee and Conversation!
DISTINGUISHED LECTURE at the ECE SEMINAR SERIES

Abstract

The need for energy efficiency in next-generation computational frameworks is imperative. One source of inspiration for new, efficient computing architectures has been the brain, which witnesses the premise that advanced intelligence systems can be built on a reasonable power budget. In this talk, I explore three recent projects from our research group that make use of brain-inspired motifs to address important problems in computing: temporal state machines and time-coded arithmetic; stochastic computing systems based on magnetic tunnel junctions; and a new architecture for neural network gradient compression based on high-dimensional rotations, inspired by unsupervised neuron models.

Bio

Dr. Matthew W. Daniels is a project leader at the National Institute of Standards and Technology (NIST) in Gaithersburg, Maryland. He received his PhD from the Physics Department at Carnegie Mellon University in 2017 under the supervision of Prof. Di Xiao, where he worked on antiferromagnetic spintronics and 2D materials. Since moving to NIST and becoming a permanent research staff member, Matthew’s research interests are in using the tools of physics to understand, develop, and quantify energy-efficient computing schemes and information encodings at the CMOS/physics interface, especially oriented toward neuromorphic and machine learning applications.

Hosted by: Distinguished Lecture at the ECE Seminar Series

Submitted by: Kerem Camsari <camsari@ucsb.edu>