NEWS

live 1.5.10 (2019-02-28)

Allow setting seed before sampling in sample_locally2 to make results reproducible.
Add new explainer: local_permutation_importance function.
Fixed problems with mlr dependency.
Add shortcut function for DALEX explainers: local_approximation.

Minor fix to euclidean_kernel function.
Default kernel in fit_explanation is now gaussian_kernel.
Order of arguments changed in add_predictions and data arguments defaults to NULL.
Variables are standardized after predictions are added, before explanation model is fitted in fit_explanation function.

Print functions for results of sample_locally, add_predictions and fit_explanation.

Observations in simulated dataset can now be weighted according to their distance from the explained instance. The distance is defined by kernel argument to fit_explanation function.
Some variables can be excluded from sampling. This is controled via fixed_variables argument to sample_locally function.
Documentation was improved.
Object returned by sample_locally, add_predictions and fit_explanation functions now carry more information (mainly explained instance) so some function calls were simplified (plot_explanation).

Variable selection is now better suited to work with factor/character variables.

fit_explanation is now more flexible, can take a list of hyperparameters for a chosen model.

For classification problems waterfall plots can be drawn on probability or logit scale.