Spinbis: Spintronics Based Bayesian Inference System with Stochastic Computing

By Xiaotao Jia et al

Published on Jan. 10, 2019

Read the original document by opening this link in a new tab.

Table of Contents

Abstract, Introduction, Analytical Macrospin Modeling of the Stochastic Switching Time of Spin-Transfer Torque Devices, Introduction, Background, Stochastic Computing, Magnetic Tunnel Junction (MTJ) Device, The SPINBIS Architecture

Summary

The document discusses the SPINBIS system, a spintronics based Bayesian inference system with stochastic computing. It proposes an efficient approach for Bayesian inference using spintronic devices. The system leverages stochastic bitstream generators and hybrid CMOS/MTJ circuits design to improve inference efficiency. Experimental results show significant improvements in energy efficiency compared to traditional approaches. The document also introduces the concept of stochastic computing, MTJ devices, and the architecture of SPINBIS.