Probabilistic Computing with Stochastic Magnetic Tunnel Junctions
ECE 595 SEMINAR SERIES
March 5, 2021
11:00 AM - 12:00 PM
Location
online: https://uic.zoom.us/j/85191358171? pwd=YWdVd2RxWE95TC9LR3JQWkMzZDltQT09
Address
Chicago, IL 60607
Calendar
Download iCal FileProbabilistic Computing with Stochastic Magnetic Tunnel Junctions
Presenter: Karem Cansari, University of California Santa Barbara
Abstract: Digital computing is based on the notion of a "bit" which is a deterministic "0" or "1". Quantum computing is based on the notion of a "qubit" which can be a superposition of "0" and "1". In this talk, attention will be drawn to something in between, which we call a probabilistic or "p-bit" which can be used as a building block for probabilistic computers--something that may be worth building.
While p-bits do not possess the computational powers of fault-tolerant ideal qubits, much of the recent applications discussed for Noisy-Intermediate Quantum (NISQ) computers are shared by probabilistic computers. These include optimization and inference problems for machine learning, computational biology, and accelerating learning algorithms, among many others.
Probabilistic bits for basic building blocks of probabilistic computers are abstractions and they admit different implementations. A particularly compact implementation comes from spintronics: The natural noise on a low-barrier nanomagnet in a magnetic tunnel junction in combination with CMOS transistors can be used to build a highly compact and energy-efficient p-bit. Our recent table-top experiment that implements 8 such interconnected MTJ/CMOS p-bits that run autonomously to solve a class of optimization problems will be discussed.
In addition, learn how such autonomously operating probabilistic computers can accelerate a class of quantum computing algorithms such as those implemented by D-Wave without facing the extreme hardware difficulties of cryogenic operation. Device and architecture level comparisons of existing implementations of p-bits will be included.
Bio: Cansari received his PhD in electrical and computer engineering from Purdue University in 2015, where he continued on as a post-doctoral researcher from 2015 to 2020, before joining the department of electrical and computer engineering at UC Santa Barbara in 2020. His PhD work established a modular approach to connect a growing set of emerging materials and phenomena to circuits and systems, a framework that has also been adopted by others. In his postdoctoral work, he used this approach to establish the concept of p-bits and p-circuits as a bridge between classical and quantum circuits to design efficient, domain-specific hardware accelerators in the new, beyond-More era of electronics.
Faculty host: Amit Trivedi, amitrt@uic.edu
This event will not be recorded.
Date posted
Mar 4, 2021
Date updated
Mar 4, 2021