fguignard/UncELMe

Python package for Uncertainty quantification of Extreme Learning Machine ensemble.

UncELMe

Python package for Uncertainty quantification of Extreme Learning Machine ensemble.

UncELMe contains :

• The ELM, ELMRidge and ELMRidgeCV classes, which are scikit-learn compatible estimators for regression based on Extreme Learning Machine (ELM), with regularization possibility (ridge estimate).

• The ELMEnsemble, ELMEnsembleRidge and ELMEnsembleRidgeCV classes, which allows ensemble of ELM, ELMRidge and ELMRidgeCV estimators.

• Estimates of model variance for the ensemble, including homoskedastic and heteroskedastic estimates for the non-regularized and regularized cases.

More theoretical and implementation details can be found in Guignard et al.. Please refer to this article if you are using the package.

License : MIT

How to install this package

The package can be installed via pip install command:

pip install UncELMe

Documentation

You can find the documentation of this repository here.

Examples

Examples with the ELMEnsemble class are availaible on GitHub Gist to help you get started :

• A one-dimensional simulated case using homoskedastic estimate.

• An example with the Boston data set using homoskedastic and heteroskedastic estimates.

Examples with ELMEnsembleRidge and ELMEnsembleRidgeCV classes will follow.