https://github.com/alisiahkoohi/srcsep

Code to partially reproduce results in "Unearthing InSights into Mars: Unsupervised source separation with limited data", ICML 2023
https://github.com/alisiahkoohi/srcsep

limited-data scattering-covariances scattering-networks source-separation unsupervised-learning wavelet

Last synced: 15 days ago
JSON representation

Code to partially reproduce results in "Unearthing InSights into Mars: Unsupervised source separation with limited data", ICML 2023

• Host: GitHub
• URL: https://github.com/alisiahkoohi/srcsep
• Owner: alisiahkoohi
• License: mit
• Created: 2023-05-30T23:33:40.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-03-11T22:41:46.000Z (8 months ago)
• Last Synced: 2024-03-11T23:40:57.798Z (8 months ago)
• Topics: limited-data, scattering-covariances, scattering-networks, source-separation, unsupervised-learning, wavelet
• Language: Python
• Homepage: https://proceedings.mlr.press/v202/siahkoohi23a.html
• Size: 108 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Unearthing InSights into Mars: unsupervised source separation with limited data


Code to partially reproduce results in [Unearthing InSights into Mars: unsupervised source separation with limited data](https://proceedings.mlr.press/v202/siahkoohi23a.html), published in the proceedings of ICML 2023.

## Installation

Run the commands below to install the required packages.

```bash

git clone https://github.com/alisiahkoohi/srcsep

cd srcsep/

conda env create -f environment.yml

conda activate srcsep

pip install -e .

```

After the above steps, you can run the example scripts by just

activating the environment, i.e., `conda activate srcsep`, the

following times.

## Scripts

Deglitching can be done for a toy example by running the following:

```bash

python scripts/toy_example.py

```

The default command line arguments are stored at `configs/toy_example.json`. Non-default arguments can be passed to the script by for example:

```bash

python scripts/toy_example.py

    --max_itr 1000 \

    --j 8,8 \

    --q 1,1 \

    --type exp_glitch

```

The generated data is stored in `data/checkpoints/` directory. To visualize the results, run:

```bash

python scripts/visualize_results.py

    --max_itr 1000 \

    --j 8,8 \

    --q 1,1 \

    --type exp_glitch

```

The figures will be stored in the `plots/` directory.

**Note regarding caching:** The scattering covariance computation caches the results in `srcsep/_cached_dir` and following runs with the same exact setup will simply load the results. Feel free to delete the cache when needed.

## Questions

Please contact [email protected] for questions.

## Authors

Rudy Morel and Ali Siahkoohi