A Probabilistic State Space Model for Joint Inference from Differential Equations and Data

Note:

Most of what makes up the method presented in the paper is implemented in ProbNum. Since ProbNum is under development and thus subject to regular changes, the paper code has yet to be adapted in parts.

Paper

Accepted (Poster) at NeurIPS 2021:

NeurIPS 2021 Proceedings (or on OpenReview ).

Please cite this work as

@inproceedings{
    schmidt2021a,
    title={A Probabilistic State Space Model for Joint Inference from Differential Equations and Data},
    author={Jonathan Schmidt and Nicholas Kr{\"a}mer and Philipp Hennig},
    booktitle={Thirty-Fifth Conference on Neural Information Processing Systems},
    year={2021},
    url={https://openreview.net/forum?id=7e4FLufwij}
}

Install

pip install -e .

Run experiments

1. Change working directory to experiment folder, e.g.

cd exp/exp003_standard_sird

2. Execute experiment, e.g.

./run_experiment ger

Plot results

(Make sure you are in the directory that contains the result directory that has just been created.)

1. Execute plotting script, e.g.

log-sird-plot <result-directory>