TY - GEN
T1 - INVERSE IMAGING WITH GENERATIVE PRIORS VIA LANGEVIN DYNAMICS
AU - Nguyen, Thanh V.
AU - Jagatap, Gauri
AU - Hegde, Chinmay
N1 - Funding Information:
Email: thanhng.cs@gmail.com, gbj221@nyu.edu, chinmay.h@nyu.edu. This work was partially done while the authors were with the Electrical and Computer Engineering Department at Iowa State University. CH is currently at the Tandon School of Engineering at New York University. This work was supported in part by NSF grants CCF-2005804 and CCF-1815101.
Funding Information:
This work was partially done while the authors were with the Electrical and Computer Engineering Department at Iowa State University. CH is currently at the Tandon School of Engineering at New York University. This work was supported in part by NSF grants CCF-2005804 and CCF-1815101.
Publisher Copyright:
© 2022 IEEE
PY - 2022
Y1 - 2022
N2 - Deep generative models have emerged as a powerful class of priors for signals in various inverse problems such as compressed sensing, phase retrieval and super-resolution. Here, we assume an unknown signal to lie in the range of some pre-trained generative model. A popular approach for signal recovery is via gradient descent in the low-dimensional latent space. While gradient descent has achieved good empirical performance, its theoretical behavior is not well understood. In this paper, we introduce the use of stochastic gradient Langevin dynamics (SGLD) for compressed sensing with a generative prior. Under mild assumptions on the generative model, we prove the convergence of SGLD to the true signal. We also demonstrate competitive empirical performance to standard gradient descent.
AB - Deep generative models have emerged as a powerful class of priors for signals in various inverse problems such as compressed sensing, phase retrieval and super-resolution. Here, we assume an unknown signal to lie in the range of some pre-trained generative model. A popular approach for signal recovery is via gradient descent in the low-dimensional latent space. While gradient descent has achieved good empirical performance, its theoretical behavior is not well understood. In this paper, we introduce the use of stochastic gradient Langevin dynamics (SGLD) for compressed sensing with a generative prior. Under mild assumptions on the generative model, we prove the convergence of SGLD to the true signal. We also demonstrate competitive empirical performance to standard gradient descent.
UR - http://www.scopus.com/inward/record.url?scp=85131252697&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85131252697&partnerID=8YFLogxK
U2 - 10.1109/ICASSP43922.2022.9747128
DO - 10.1109/ICASSP43922.2022.9747128
M3 - Conference contribution
AN - SCOPUS:85131252697
T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
SP - 8672
EP - 8676
BT - 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022
Y2 - 23 May 2022 through 27 May 2022
ER -