# Generative Model with Dynamic Linear Flow

## TL;DR

The paper introduces Dynamic Linear Flow, a novel invertible transformation that combines the advantages of flow-based and autoregressive models, achieving state-of-the-art density estimation and faster convergence.

## Contribution

It proposes Dynamic Linear Flow, a new invertible transformation with a partially autoregressive structure, improving density estimation and convergence speed over existing flow models.

## Key findings

- State-of-the-art performance on ImageNet 32x32 and 64x64 among flow-based models.
- Converges 10 times faster than Glow.
- Competitive with top autoregressive models.

## Abstract

Flow-based generative models are a family of exact log-likelihood models with tractable sampling and latent-variable inference, hence conceptually attractive for modeling complex distributions. However, flow-based models are limited by density estimation performance issues as compared to state-of-the-art autoregressive models. Autoregressive models, which also belong to the family of likelihood-based methods, however suffer from limited parallelizability. In this paper, we propose Dynamic Linear Flow (DLF), a new family of invertible transformations with partially autoregressive structure. Our method benefits from the efficient computation of flow-based methods and high density estimation performance of autoregressive methods. We demonstrate that the proposed DLF yields state-of-theart performance on ImageNet 32x32 and 64x64 out of all flow-based methods, and is competitive with the best autoregressive model. Additionally, our model converges 10 times faster than Glow (Kingma and Dhariwal, 2018). The code is available at https://github.com/naturomics/DLF.