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.
live 1.5.7
New method of sampling (“normal”).
1ive 1.5.6
Waterfall plots can be viewed in a Shiny app.
live 1.5.5
Fixed bug related to standardizing columns in
fit_explanation.
live 1.5.4
Old interface dropped.
live 1.5.3
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.
live 1.5.2
Print functions for results of sample_locally, add_predictions and
fit_explanation.
live 1.5.1
New, LIME-like method of sampling as an option in
sample_locally.
live 1.5.0
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).
live 1.4.2
Fixed bug in variable selection.
live 1.4.1
Variable selection is now better suited to work with
factor/character variables.
live 1.4.0
Variable selection is now based on LASSO as implemented in glmnet
package.
Updated documentation and vignette.
live 1.3.3
add_predictions also returns black box model object
(model element).
live 1.3.2
Hyperparameters can be also passed to add_predictions
function.
live 1.3.1
fit_explanation is now more flexible, can take a list
of hyperparameters for a chosen model.
live 1.3.0
For classification problems waterfall plots can be drawn on
probability or logit scale.
live 1.2.0
Now using forestmodel package for better factor handling.
live 1.1.2
Date variables will now be hold constant while performing local
exploration.
Improved performance.
live 1.1.1
add_predictions improved to handle more learners (for
example ranger).
live 1.1.0
Added a NEWS.md file to track changes to the
package.
sample\_locally uses data.table for faster local
exploration.