My research develops principled computational frameworks that bridge physics-based models, statistical inference, and machine learning.
Plug-and-Play Priors (PnP)
The Plug-and-Play Priors framework enables flexible integration of powerful denoising algorithms as Bayesian priors in inverse problems. Adopted by 100+ research laboratories worldwide and implemented in Samsung Galaxy smartphone fingerprint sensors.
Consensus Equilibrium
Multi-agent optimization framework for modular integration of disparate sources of expertise. Led to a $1.2M NSF research grant.
Latent Space Geometry for Explainable AI
Using differential geometric tools (Fisher-Rao metric) to diagnose failures in representation learning networks.
Privacy-Preserving Machine Learning
Privacy-preserving face identification using generative neural networks. Awarded Wells Fargo Medallion for Most Innovative Patent (2020).
Selected Publications
Plug-and-play unplugged: Optimization-free reconstruction using consensus equilibrium
G.T. Buzzard, S.H. Chan, S. Sreehari, C.A. Bouman
SIAM Journal on Imaging Sciences, 11(3), 2001-2020, 2018