https://github.com/borgwardtlab/topo-ae-distances
https://github.com/borgwardtlab/topo-ae-distances
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/borgwardtlab/topo-ae-distances
- Owner: BorgwardtLab
- License: bsd-3-clause
- Created: 2020-10-02T15:00:26.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2021-09-08T05:59:35.000Z (over 4 years ago)
- Last Synced: 2025-07-11T11:54:38.359Z (11 months ago)
- Language: Python
- Size: 50.6 MB
- Stars: 8
- Watchers: 4
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Challenging Euclidean Topological Autoencoders
This is a follow-up project of the ICML 2020 paper "Topological Autoencoders" (reference below).
Here, we investigate whether domain-specific distance functions in the input space (here image datasets) are necessary for TopoAE, or whether a generic euclidean distance is sufficient.
[This work](https://openreview.net/pdf?id=P3dZuOUnyEY) has been accepted for presentation at the Neurips 2020 TDA and Beyond workshop.
## References
Please use the following BibTex code to cite our Neurips 2020 workshop [paper](https://openreview.net/pdf?id=P3dZuOUnyEY):
```
@InProceedings{moor2020challenging,
title = {Challenging Euclidean Topological Autoencoders},
author = {Moor, Michael and Horn, Max and Borgwardt, Karsten and Rieck, Bastian},
booktitle = {NeurIPS 2020 Workshop on Topological Data Analysis and Beyond},
year = {2020},
url = {https://openreview.net/forum?id=P3dZuOUnyEY},
}
```
Furthermore, the original ICML 2020 [paper](https://arxiv.org/abs/1906.00722) proposing Topological Autoencoders in the first place, can be cited as follows:
```
@InProceedings{Moor19Topological,
author = {Moor, Michael and Horn, Max and Rieck, Bastian and Borgwardt, Karsten},
title = {Topological Autoencoders},
year = {2020},
eprint = {1906.00722},
archiveprefix = {arXiv},
primaryclass = {cs.LG},
booktitle = {Proceedings of the 37th International Conference on Machine Learning~(ICML)},
series = {Proceedings of Machine Learning Research},
publisher = {PMLR},
pubstate = {forthcoming},
}
```
## Setup
In order to reproduce the results indicated in the workshop paper simply setup an
environment using poetry:
```bash
poetry install
```
## Running the methods:
Make sure you have internet access once to be able to download the datasets, and also the vgg model (via the lpips package)
In case a slurm cluster is available, simply run:
```bash
source scripts/run_slurm.sh
```
Alternatively, all jobs can be sequentially/manually called using:
```bash
source scripts/run_manual.sh
```