GPyConform

GPyConform extends the GPyTorch library by implementing (Full) Conformal Prediction for Gaussian Process Regression based on the approach described in [1]. Designed to work seamlessly with Exact Gaussian Process (GP) models, GPyConform enhances GPyTorch by introducing the capability to generate and evaluate both 'symmetric' and 'asymmetric' Conformal Prediction Intervals.

Key Features

Provides Provably Valid Prediction Intervals: Provides Prediction Intervals with guaranteed coverage under minimal assumptions (data exchangeability).
Full Utilization of GPyTorch: Leverages the robust and efficient GP modeling capabilities of GPyTorch.
Supports Both Symmetric and Asymmetric Prediction Intervals: Implements both the symmetric and asymmetric Full Conformal Prediction approaches for constructing Prediction Intervals.

Note

Currently, GPyConform is tailored specifically for Exact GP models combined with any covariance function that employs an exact prediction strategy.

Documentation

For detailed documentation and usage examples, see GPyConform Documentation.

Installation

From PyPI

pip install gpyconform

From conda-forge

conda install conda-forge::gpyconform

Citing GPyConform

If you use GPyConform for a scientific publication, you are kindly requested to cite the following paper:

Harris Papadopoulos. Guaranteed Coverage Prediction Intervals with Gaussian Process Regression. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. DOI: 10.1109/TPAMI.2024.3418214. (arXiv version)

Bibtex entry:

@ARTICLE{gprcp,
  author={Papadopoulos, Harris},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  title={Guaranteed Coverage Prediction Intervals with Gaussian Process Regression}, 
  year={2024},
  volume={},
  number={},
  pages={1-12},
  doi={10.1109/TPAMI.2024.3418214}
}

References

[1] Harris Papadopoulos. Guaranteed Coverage Prediction Intervals with Gaussian Process Regression. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. DOI: 10.1109/TPAMI.2024.3418214. (arXiv version)

[2] Vladimir Vovk, Alexander Gammerman, and Glenn Shafer. Algorithmic Learning in a Random World, 2nd Ed. Springer, 2023. DOI: 10.1007/978-3-031-06649-8.