Evaluate the performance of (probabilistic) numerical methods. What makes an efficient and well calibrated probabilistic numerical algorithm? We don't know either, but at least we can offer some quantification.
ProbNum-Evaluation offers a range of evaluation measures, including error analyses (with respect to different norms), calibration measures (chi-square statistics, non-credibility indices) and more.
To learn how to use ProbNum-Evaluation check out the quickstart guide and the tutorials. For guidelines how to contribute to ProbNum-Evaluation, please refer to the ProbNum documentation <https://probnum.readthedocs.io/en/latest/development/contributing.html>.
public_api/timeseries public_api/multivariate public_api/visual public_api/config
tutorials/measure_error_and_calibration tutorials/animate_gaussian_distributions tutorials/animate_gauss_markov_posterior
GitHub Repository <https://github.com/probabilistic-numerics/probnum-evaluation> license
genindexmodindex