https://github.com/rpatrik96/ima-vae

This is the code for the paper Embrace the Gap: VAEs perform Independent Mechanism Analysis, showing that optimizing the ELBO is equivalent to optimizing the IMA-regularized log-likelihood under certain assumptions (e.g., small decoder variance).
https://github.com/rpatrik96/ima-vae

disentanglement dsprites ica independent-component-analysis pytorch pytorch-lightning vae variational-autoencoder variational-inference wandb

Last synced: 12 months ago
JSON representation

This is the code for the paper Embrace the Gap: VAEs perform Independent Mechanism Analysis, showing that optimizing the ELBO is equivalent to optimizing the IMA-regularized log-likelihood under certain assumptions (e.g., small decoder variance).

• Host: GitHub
• URL: https://github.com/rpatrik96/ima-vae
• Owner: rpatrik96
• License: apache-2.0
• Created: 2021-11-25T10:55:19.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2024-04-22T13:43:43.000Z (about 2 years ago)
• Last Synced: 2025-06-02T10:44:26.318Z (about 1 year ago)
• Topics: disentanglement, dsprites, ica, independent-component-analysis, pytorch, pytorch-lightning, vae, variational-autoencoder, variational-inference, wandb
• Language: Jupyter Notebook
• Homepage:
• Size: 11.4 MB
• Stars: 23
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE
  • Citation: CITATION.cff

Awesome Lists containing this project

README

          
---

    

 

# Embrace the Gap: VAEs perform Independent Mechanism Analysis   


[//]: # ([![Paper](http://img.shields.io/badge/paper-arxiv.2206.02416-B31B1B.svg)](https://arxiv.org/abs/2206.02416))

[![Conference](http://img.shields.io/badge/NeurIPS-2022-4b44ce.svg)](https://openreview.net/forum?id=G4GpqX4bKAH)

[//]: # ([![Conference](http://img.shields.io/badge/ICLR-2019-4b44ce.svg)](https://papers.nips.cc/book/advances-in-neural-information-processing-systems-31-2018))

[//]: # ([![Conference](http://img.shields.io/badge/AnyConference-year-4b44ce.svg)](https://papers.nips.cc/book/advances-in-neural-information-processing-systems-31-2018)  )

[![Paper](http://img.shields.io/badge/arxiv-stat.ML:2206.02416-B31B1B.svg)](https://arxiv.org/abs/2206.02416)

![CI testing](https://github.com/rpatrik96/ima-vae/workflows/CI%20testing/badge.svg?branch=master&event=push)

[![DOI](https://zenodo.org/badge/431811003.svg)](https://zenodo.org/badge/latestdoi/431811003)

   



 

## Description   

This is the code for the paper _Embrace the Gap: VAEs perform Independent Mechanism Analysis_, showing that optimizing the ELBO is equivalent to optimizing the IMA-regularized log-likelihood under certain assumptions (e.g., small decoder variance).


## How to run   

First, install dependencies   

```bash

# clone ima_vae   

git clone --recurse-submodules https://github.com/rpatrik96/ima-vae

# if forgot to pull submodules, run

git submodule update --init

# install ima_vae   

cd ima-vae

pip install -e .   

pip install -r requirements.txt

# install spriteworld

pip install -e ./spriteworld

# install submodule requirements

pip install --requirement ima/requirements.txt --quiet

pip install --requirement tests/requirements.txt --quiet

pip install --requirement spriteworld/requirements.txt --quiet

# install pre-commit hooks (only necessary for development)

pre-commit install

 ```   

 Next, navigate to the `ima-vae` directory and run `ima_vae/cli.py.   

```bash

 python3 ima_vae/cli.py fit --help

 python3 ima_vae/cli.py fit --config configs/trainer.yaml --config configs/synth/moebius.yaml --model.prior=beta

```

### Hyperparameter optimization

First, you need to log into `wandb`

```bash

wandb login #you will find your API key at https://wandb.ai/authorize

```

Then you can create and run the sweep

```bash

wandb sweep sweeps/synth/mlp/finding_optimal_gamma_uniform.yaml  # returns sweep ID

wandb agent  --count= # when used on a cluster, set it to one and start multiple processes

```

## Citation   

```

@inproceedings{

 reizinger_embrace_2022,

 title={Embrace the Gap: {VAE}s Perform Independent Mechanism Analysis},

 author={Patrik Reizinger and Luigi Gresele and Jack Brady and Julius Von K{\"u}gelgen and Dominik Zietlow and Bernhard Sch{\"o}lkopf and Georg Martius and Wieland Brendel and Michel Besserve},

 booktitle={Advances in Neural Information Processing Systems},

 editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},

 year={2022},

 url={https://openreview.net/forum?id=G4GpqX4bKAH}

}

```