This repository contains the PyTorch implementation of our paper: MintNet: Building Invertible Neural Networks with Masked Convolutions, NeurIPS 2019 . We propose a new way of constructing invertible neural networks by combining simple building blocks with a novel set of composition rules. This leads to a rich set of invertible architectures, including those similar to ResNets. Inversion is achieved with a locally convergent iterative procedure that is parallelizable and very fast in practice. Additionally, the determinant of the Jacobian can be computed analytically and efficiently, enabling their generative use as flow models.

The following are packages needed for running this repo.

• PyTorch==1.1.0
• tqdm
• tensorboardX
• Scipy
• PyYAML
• Numba

python main.py --runner [runner name] --config [config file] --doc [experiment folder name]

Here runner name is one of the following:

• DensityEstimationRunner. Experiments on MintNet density estimation.
• ClassificationRunner. Experiments on MintNet classification.

config file is the directory of some YAML file in configs/, and experiment folder name is the folder names in run/.

For example, if you want to train MintNet density estimation model on MNIST, just run

python main.py --runner DensityEstimationRunner --config mnist_density_config.yml

Checkpoints for both density estimation and classification can be downloaded from https://drive.google.com/file/d/12kGMMg0ivJI5y32hRouhZuddr9cJxfiR/view?usp=sharing

Unzip it to <root folder>/run.