Skip to yearly menu bar Skip to main content

Workshop: Machine Learning and the Physical Sciences

GAN-Flow: A dimension-reduced variational framework for physics-based inverse problems

Agnimitra Dasgupta · Dhruv Patel · Deep Ray · Erik Johnson · Assad Oberai


We propose GAN-Flow -- a modular inference approach that combines generative adversarial network (GAN) prior with a normalizing flow (NF) model to solve inverse problems in the lower-dimensional latent space of the GAN prior using variational inference. GAN-Flow leverages the intrinsic dimension reduction and superior sample generation capabilities of GANs, and the capability of NFs to efficiently approximate complicated posterior distributions. In this work, we apply GAN-Flow to solve two physics-based linear inverse problems. Results show that GAN-Flow can efficiently approximate the posterior distribution in such high-dimensional problems.

Chat is not available.