Oct 20 (Tue) @ 12:30pm: "Energy-Based Probabilistic Computing: From Programmable Spin Hardware to Learning-Augmented Monte Carlo," Saleh Bunaiyan, ECE PhD Defense

Location: Engineering Science Bldg (ESB), Room 1002
Zoom Link: https://ucsb.zoom.us/j/5025931709

Abstract

Probabilistic computing has emerged as a promising energy-efficient approach to solving complex optimization and sampling problems using networks of probabilistic bits (p-bits). This dissertation explores two directions for advancing probabilistic computing: extending binary p-bits to represent more complex systems and accelerating Monte Carlo sampling using machine learning.

First, we investigate a spintronic architecture for extending conventional binary p-bits to continuous, vector-valued spins. Specifically, we propose a device that enables sampling from the Heisenberg model, where each spin is represented by a unit vector on a sphere rather than a binary state, as in the Ising model. Second, we introduce Transformer-Augmented Parallel Tempering (TAPT), a framework that combines generative models with conventional Monte Carlo methods. We develop IsingFormer, a Transformer trained on Ising configurations to learn their statistical correlations and generate nonlocal proposals that complement conventional Monte Carlo updates. These proposals help the system explore complex energy landscapes more efficiently, improving sampling and optimization performance in problems such as spin glasses and integer factorization.

Bio

Saleh Bunaiyan is a Ph.D. candidate in Electrical and Computer Engineering at the University of California, Santa Barbara, advised by Professor Kerem Y. Camsari. His research focuses on energy-based probabilistic computing, spanning spintronic hardware and learning-augmented Monte Carlo methods. His work explores spintronic architectures for Heisenberg-inspired probabilistic computing and generative models for accelerating sampling and optimization in Ising systems.

Prior to his doctoral studies, he earned a Master of Science in Electrical Engineering (2022) and a Bachelor of Science in Electrical Engineering (2020), both from King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia. During his master's studies, he conducted research at the Center for Energy and Geo-Processing (CeGP), a collaboration between KFUPM and the Georgia Institute of Technology. He also completed a research internship at King Abdullah University of Science and Technology (KAUST) in 2019 under Professor Muhammad Mustafa Hussain.

Hosted By: ECE PhD Defense

Submitted By: Saleh Bunaiyan | Email