Today, in NeurONN Lecture Series, we welcome Prof. Kerem Camsari, UC Santa Barbara. His talk will be on Massively Parallel Probabilistic Computing with p-bits

Title: Massively Parallel Probabilistic Computing with p-bits

Abstract: The slowing down of Moore’s Law has led to a crisis as the computing workloads of Artificial Intelligence (AI) algorithms continue skyrocketing. There is an urgent need for scalable, energy-efficient hardware catering to the unique requirements of AI algorithms and applications. In this environment, probabilistic computing (p-computing) with p-bits [1] has emerged as a scalable, domain-specific and energy-efficient computing paradigm, particularly useful for probabilistic applications and algorithms.

In this talk, I will describe two general applications of p-computing: optimization and learning as problems relevant to Machine Learning and AI. I will talk about recent and representative [2-3] experiments illustrating how both problems can be efficiently addressed by a suitably modified magnetoresistive random access memory (MRAM) technology. I will then show standard silicon-based implementations of p-computing applied to practical optimization problems in large scale [3] to stress why nanodevice-based implementations of p-computing is a crucially needed ingredient.

[1] Camsari, Kerem, et al. “Stochastic p-bits for invertible logic.” Physical Review X (2017)

[2] Borders, William A., et al. “Integer factorization using stochastic magnetic tunnel junctions.” Nature (2019)

[3] Aadit, Navid Anjum, et al. “Massively Parallel Probabilistic Computing with Sparse Ising Machines.” Nature Electronics (2022)