crepes 0.0.1

Conformal regressors and predictive systems (Crepes)

crepes

Conformal regressors and predictive systems

Install with:

pip install crepes

The package implements conformal regressors, which transform point predictions (produced by any underlying regression model) into prediction intervals, for specified levels of confidence.

The package also implements conformal predictive systems, which transform the point predictions into cumulative distributions (conformal predictive distributions), e.g., allowing prediction intervals to be extracted as well as probabilities for the target value falling below specified thresholds.

The package implements standard, normalized and Mondrian conformal regressors and predictive systems, and allows for using both built-in and tailored difficulty estimates and Mondrian categories.

For examples of how to use the package, see this Jupyter notebook.

Author: Henrik Boström ([e-mail]) Copyright 2021 Henrik Boström License: BSD 3 clause