Repository for paper "Unsupervised Disentanglement without Autoencoding: Pitfalls and Future Directions" featured at the ICML 2021 Workshop Self-Supervised Learning for Reasoning and Perception. Paper can be found here, as well as our workshop poster here.
We provide the colab used to generate the MNIST/STL-10 dataset used for all our disentanglement experiments and the resulting data files for use by others.
If you make use of the paper or the MNIST/STL-10 data, please cite our paper:
@article{burns2021unsupervised,
title={Unsupervised Disentanglement without Autoencoding: Pitfalls and Future Directions},
author={Andrea Burns and Aaron Sarna and Dilip Krishnan and Aaron Maschinot},
year={2021},
journal={arXiv preprint arXiv:2108.06613}
}