A Method to Model Conditional Distributions with Normalizing Flows

TL;DR

This paper introduces a streamlined method using normalizing flows for modeling conditional distributions, simplifying training and enabling effective inverse problem analysis and conditional generation.

Contribution

The proposed approach simplifies training by using a single loss, improves upon previous methods, and provides a natural framework for conditional generation with invertible neural networks.

Findings

Effective in modeling conditional distributions

Simplifies training process

Enables conditional generation

Abstract

In this work, we investigate the use of normalizing flows to model conditional distributions. In particular, we use our proposed method to analyze inverse problems with invertible neural networks by maximizing the posterior likelihood. Our method uses only a single loss and is easy to train. This is an improvement on the previous method that solves similar inverse problems with invertible neural networks but which involves a combination of several loss terms with ad-hoc weighting. In addition, our method provides a natural framework to incorporate conditioning in normalizing flows, and therefore, we can train an invertible network to perform conditional generation. We analyze our method and perform a careful comparison with previous approaches. Simple experiments show the effectiveness of our method, and more comprehensive experimental evaluations are undergoing.