Backlog

[interpret-ml]

This issue is to maintain links to all feature requests and bugs in one place. Issues listed here should have either the "enhancement " or "bug" label applied to them.

Issues not listed here are searchable with: is:issue is:open -label:bug -label:enhancement

Enhancements, python

• add "warm_start" parameter to enable stage customization during fitting (https://github.com/interpretml/interpret/issues/304). Also see issue # 403
• add a prediction function that also calculates uncertainty in the prediction (https://github.com/interpretml/interpret/issues/235)
• avoid resetting the index on call to show after explain_local (https://github.com/interpretml/interpret/issues/72)
• user defined labels for local explanation samples (https://github.com/interpretml/interpret/issues/300)
• change boosting algorithm to stop it from producing graphs with large positive or negative scores in some regions (https://github.com/interpretml/interpret/issues/122)
• ranking (https://github.com/interpretml/interpret/issues/181)
• Dask arrays for large datasets (https://github.com/interpretml/interpret/issues/249)
• progress indication during training (https://github.com/interpretml/interpret/issues/252)
• continuous output of metrics during training (https://github.com/interpretml/interpret/issues/7)
• Spyder support for show function (https://github.com/interpretml/interpret/issues/191)
• UI to show missing value scores (https://github.com/interpretml/interpret/issues/18)
• show error bars for interactions in the UI, perhaps with a hover (https://github.com/interpretml/interpret/issues/127)
• Improve contrast of interaction heatmaps (https://github.com/interpretml/interpret/issues/236)
• dashboard data disappears when changing tabs (https://github.com/interpretml/interpret/issues/106)
• add permutation Importance (https://github.com/interpretml/interpret/issues/26)
• alternative feature importance techniques (https://github.com/interpretml/interpret/issues/374)
• add categorical support to Morris Sensitivity Analysis (https://github.com/interpretml/interpret/issues/275)
• alternate base estimators such as linear models, polynomial regression, or splines (https://github.com/interpretml/interpret/issues/185)

Enhancements, R

• local explanations in R (https://github.com/interpretml/interpret/issues/222)
• R multiple new features (https://github.com/interpretml/interpret/issues/294)
• add threading to R and in general move the parallelism into C++ (https://github.com/interpretml/interpret/issues/180)

Bugs

• predictions in R different from python (https://github.com/interpretml/interpret/issues/417)
• cannot create multiple visualizations in Databricks (https://github.com/interpretml/interpret/issues/88)
• DataException: MissingColumnsInData, Expected column(s) 0 not found in fitted data (https://github.com/interpretml/interpret/issues/209)
• RandomizedSearchCV doesn't work with InterpretML LogisticRegression (https://github.com/interpretml/interpret/issues/301)
• categorical visualize bug (https://github.com/interpretml/interpret/issues/119)
• lime results may differ between InterpretML and LIME package (https://github.com/interpretml/interpret/issues/133)
• multiclass LIME (https://github.com/interpretml/interpret/issues/306)
• LimeTabular seems to require Pandas dataframe for categoricals (https://github.com/interpretml/interpret/issues/322)
• Categorical features in PartialDependence (https://github.com/interpretml/interpret/issues/217)
• show does not display any output in some settings (https://github.com/interpretml/interpret/issues/229)
• ClassHistogram().explain_data() having index error (https://github.com/interpretml/interpret/issues/274)
• raise an exception when the multi-model dashboard is used inside the InlineProvider (https://github.com/interpretml/interpret/issues/333)
• in some scenarios showing multiple EBM dashboards could fail (https://github.com/interpretml/interpret/issues/350)