10:45 AM - 11:15 AM
[B-6-01 (Invited)] Hardware Implementation of Probabilistic Models through Nanodevice Variability
This paper investigates the utilization of nanodevices as physical random variables to enable low-power hardware implementations of probabilistic models. Specifically, two approaches are explored: the first approach employs resistive memories as multi-level random variables for implementing Bayesian Neural Networks. The second approach involves Ising Networks using stochastic magnetic tunnel junctions to spontaneously tunable random bit streams.
