https://github.com/nicola-decao/s-vae-pytorch

Pytorch implementation of Hyperspherical Variational Auto-Encoders
https://github.com/nicola-decao/s-vae-pytorch

deep-learning hyperspherical-vae machine-learning manifold-learning pytorch vae variational-autoencoder von-mises-fisher

Last synced: 3 months ago
JSON representation

Pytorch implementation of Hyperspherical Variational Auto-Encoders

• Host: GitHub
• URL: https://github.com/nicola-decao/s-vae-pytorch
• Owner: nicola-decao
• License: mit
• Created: 2018-06-29T07:52:11.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2020-03-21T16:06:50.000Z (over 4 years ago)
• Last Synced: 2024-07-04T02:13:51.404Z (4 months ago)
• Topics: deep-learning, hyperspherical-vae, machine-learning, manifold-learning, pytorch, vae, variational-autoencoder, von-mises-fisher
• Language: Python
• Homepage: http://arxiv.org/abs/1804.00891
• Size: 27.3 KB
• Stars: 341
• Watchers: 7
• Forks: 58
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Hyperspherical Variational Auto-Encoders

### Pytorch implementation of Hyperspherical Variational Auto-Encoders

## Overview

This library contains a Pytorch implementation of the hyperspherical variational auto-encoder, or S-VAE, as presented in [[1]](#citation)(http://arxiv.org/abs/1804.00891). Check also our blogpost (https://nicola-decao.github.io/s-vae).

* Don't use Pytorch? Take a look [here](https://github.com/nicola-decao/s-vae-tf) for a **tensorflow** implementation!

## Dependencies

* **python>=3.6**

* **pytorch>=0.4.1**: https://pytorch.org

* **scipy**: https://scipy.org

* **numpy**: https://www.numpy.org

## Installation

To install, run

```bash

$ python setup.py install

```

## Structure

* [distributions](https://github.com/nicola-decao/s-vae-pytorch/tree/master/hyperspherical_vae/distributions): Pytorch implementation of the von Mises-Fisher and hyperspherical Uniform distributions. Both inherit from `torch.distributions.Distribution`.

* [ops](https://github.com/nicola-decao/s-vae-pytorch/tree/master/hyperspherical_vae/ops): Low-level operations used for computing the exponentially scaled modified Bessel function of the first kind and its derivative.

* [examples](https://github.com/nicola-decao/s-vae-pytorch/tree/master/examples): Example code for using the library within a PyTorch project.

## Usage

Please have a look into the [examples folder](https://github.com/nicola-decao/s-vae-pytorch/tree/master/examples). We adapted our implementation to follow the structure of the [Pytorch probability distributions](https://pytorch.org/docs/stable/distributions.html).

Please cite [[1](#citation)] in your work when using this library in your experiments.

## Sampling von Mises-Fisher

To sample the von Mises-Fisher distribution we follow the rejection sampling procedure as outlined by [Ulrich, 1984](http://www.jstor.org/stable/2347441?seq=1#page_scan_tab_contents). This simulation pipeline is visualized below:



  



_Note that as ![](http://latex.codecogs.com/svg.latex?%5Comega) is a scalar, this approach does not suffer from the curse of dimensionality. For the final transformation, ![](http://latex.codecogs.com/svg.latex?U%28%5Cmathbf%7Bz%7D%27%3B%5Cmu%29), a [Householder reflection](https://en.wikipedia.org/wiki/Householder_transformation) is utilized._

## Feedback

For questions and comments, feel free to contact [Nicola De Cao](mailto:[email protected]) or [Tim Davidson](mailto:[email protected]).

## License

MIT

## Citation

```

[1] Davidson, T. R., Falorsi, L., De Cao, N., Kipf, T.,

and Tomczak, J. M. (2018). Hyperspherical Variational

Auto-Encoders. 34th Conference on Uncertainty in Artificial Intelligence (UAI-18).

```

BibTeX format:

```

@article{s-vae18,

  title={Hyperspherical Variational Auto-Encoders},

  author={Davidson, Tim R. and

          Falorsi, Luca and

          De Cao, Nicola and

          Kipf, Thomas and

          Tomczak, Jakub M.},

  journal={34th Conference on Uncertainty in Artificial Intelligence (UAI-18)},

  year={2018}

}

```