Dr. Andrew Holbrook, Associate Professor of Biostatistics at the UCLA Fielding School of Public Health, will be speaking about "Multiproposal MCMC from GPUs to Quantum Circuits" on Wednesday, October 28th, 2026 from 3:30 - 4:30pm in HSSB 1173.
Abstract:
Multiproposal MCMC algorithms generate a cloud of proposals at each iteration and select the next state among them, offering parallelism distinct from running many chains. This talk surveys their foundations, design and acceleration: an involutive framework that characterizes reversibility on general state spaces and yields symmetric proposal structures; the infinite proposal limit, which exposes degenerate schemes and uncovers the slingshot, an algorithm biased for finite proposal counts yet asymptotically exact; and quantum algorithms that cut target evaluations per step from O(P) to O(√P) to O(1). I close with QAOA-based quantum proposals and Stein variational gradient descent proposals with slingshot acceptance.
Bio:
Andrew J. Holbrook, Ph.D., is an Associate Professor of Biostatistics at the UCLA Fielding School of Public Health, an NSF CAREER Awardee, and an NIGMS MIRA Awardee. He has contributed to two popular open-source scientific software platforms: ANTsX for quantitative medical imaging and BEAST for Bayesian phylogenetic inference. He builds Bayesian statistical and machine learning methods for the largest problems in health, including viral epidemics, Alzheimer’s disease, gun violence, and glaucoma. His work includes parallel algorithms that track global viral spread, generative AI models for phylogenetics, interpretable AI for neural decoding, and quantum algorithms for Bayesian inference.