Stochastic Spin-Orbit Torque Devices as Elements for Bayesian Inference

Abstract

Probabilistic inference from real-time input data is becoming increasingly popular and may be one of the potential pathways at enabling cognitive intelligence. As a matter of fact, preliminary research has revealed that stochastic functionalities also underlie the spiking behavior of neurons in cortical microcircuits of the human brain. In tune with such observations, neuromorphic and other unconventional computing platforms have recently started adopting the usage of computational units that generate outputs probabilistically, depending on the magnitude of the input stimulus. In this work, we experimentally demonstrate a spintronic device that offers a direct mapping to the functionality of such a controllable stochastic switching element. We show that the probabilistic switching of Ta/CoFeB/MgO heterostructures in presence of spin-orbit torque and thermal noise can be harnessed to enable probabilistic inference in a plethora of unconventional computing scenarios. This work can potentially pave the way for hardware that directly mimics the computational units of Bayesian inference.

Stochastic Memory Devices for Security and Computing

“…[164] Another concern is that A and B should be statistically independent in Equation (3). [172] The device structure, shown in Figure 16c, comprises a CoFeB/MgO/Ta MTJ built on top of a 10 nm-thick Ta layer. However, the availability of TRNG in stochastic memory devices provides the necessary stochastic bit stream within a small area and low power consumption.…”

“…[173][174][175] Here, each node represents random variables, while each link describes the direct dependence among variables. [172] The dependence between each variable is quantified with the conditional probabilities of a transition to a particular node from its parent node, as described by the conditional probability table (CPT). [172] The dependence between each variable is quantified with the conditional probabilities of a transition to a particular node from its parent node, as described by the conditional probability table (CPT).…”

“…[173] The BN in the figure include four variables, namely the probability of being "cloudy (C)," "rainy (R)," having the "sprinkler on (S)" or the grass being "wet (W)." [172] This BN can be used in a probabilistic inference engine to estimate the probability of hidden causes from a given observation. [172] This BN can be used in a probabilistic inference engine to estimate the probability of hidden causes from a given observation.…”

“…[172] The dependence between each variable is quantified with the conditional probabilities of a transition to a particular node from its parent node, as described by the conditional probability table (CPT). [172] For instance, the observation could be wet grass in Figure 16d, with the objective of estimating the cause, whether the sprinkler is on, or it is raining. [172] For instance, the observation could be wet grass in Figure 16d, with the objective of estimating the cause, whether the sprinkler is on, or it is raining.…”

“…A Bayesian inference engine must clearly implement Bayes' rule of [172] Published 2017, Springer Nature. A Bayesian inference engine must clearly implement Bayes' rule of [172] Published 2017, Springer Nature.…”