Interpretable t-SNE

interpretable_tsne is an implementation of our gradient-based attributiont technique described in our GLBIO oral presentation: 'Towards Computing Attributions for Dimensionality Reduction Techniques'.

To replicate the experiments we performed in our presentation, go to: MattScicluna/interpretable_tsne_experiment

Synthetic data experiment demonstrating that our t-SNE attributions can identify the ground truth significant features. See our pre-print for more details.

tSNE_vid.mp4

---

Installation

interpretable_tsne requires Python 3.8 or higher to run. It is only available on Linux operating systems.

PyPi

This code can be installed via pip using the following command:

pip install interpretable_tsne

Source

You can install interpretable_tsne from source. Just clone this repository and run the following line in the root directory:

pip install .

Run the unittests

python -m unittest -v tests/*.py

---

Citation

If you use interpretable_tsne in your work, please cite us:

    @article {Scicluna2023.05.12.540592,
        author = {Matthew Crispin Scicluna and Jean-Christophe Grenier and Raphael Poujol and Sebastien Lemieux and Julie Hussin},
        title = {Towards Computing Attributions for Dimensionality Reduction Techniques},
        elocation-id = {2023.05.12.540592},
        year = {2023},
        doi = {10.1101/2023.05.12.540592},
        publisher = {Cold Spring Harbor Laboratory},
        URL = {https://www.biorxiv.org/content/early/2023/05/14/2023.05.12.540592},
        eprint = {https://www.biorxiv.org/content/early/2023/05/14/2023.05.12.540592.full.pdf},
        journal = {bioRxiv}
    }