HopCPT - Conformal Prediction for Time Series with Modern Hopfield Networks

Andreas Auer¹, Martin Gauch^1,2, Daniel Klotz¹, Sepp Hochreiter¹

¹ELLIS Unit Linz and LIT AI Lab, Institute for Machine Learning, Johannes Kepler University Linz, Austria
²Google Research, Linz, Austria\

This repository contains the source code for "Conformal Prediction for Time Series with Modern Hopfield Networks" accepted at the at Neurips 2023. The paper is available here.

Blogpost comming soon - stay tuned!

Install Dependencies

conda env update -n <your-enviroment> --file ./conformal-conda-env.yaml
pip install -r ./conformal-pip-requirements.txt
pip install neuralhydrology

Reproduce Experiments

Neuips 2023 - "Conformal Prediction for Time Series with Modern Hopfield Networks"

To re-run the experiments of Conformal Prediction for Time Series with Modern Hopfield Networks see experiments_neurips23.md.

📚 Cite

If you find this work helpful, please cite

@inproceedings{auer2023conformal,
    author={Auer, Andreas and Gauch, Martin and Klotz, Daniel and Hochreiter, Sepp},
    title={Conformal Prediction for Time Series with Modern Hopfield Networks},
    booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
    institution = {Institute for Machine Learning, Johannes Kepler University, Linz},
    year={2023},
    url={https://openreview.net/forum?id=KTRwpWCMsC}
}

Keywords

Time Series, Uncertainty, Conformal Prediction, Machine Learning, Deep Learning,