How interpretml/interpret works
InterpretML is a machine-learning interpretability package: users fit interpretable glassbox models or explain blackbox predictors, inspect global or per-prediction explanations, and visualize or compare them. The documented EBM workflow includes fitting, explanation generation, and display; other README-listed techniques form parallel model and explainer branches. EBM relies on shared native computation, while privacy-preserving EBMs and performance measures are additional capabilities. The graph follows the README and repository paths; sampled source excerpts are partial, so uncertain internal call wiring is omitted.
Main entry points: _explanation.py Explanation Objects · _ebm.py Explainable Boosting · blackbox Blackbox Explainers · greybox Greybox Explainers
Components 17
Glassbox Models
- Explainable Boosting
python/interpret-core/interpret/glassbox/_ebm.py - Other Glassbox Models
python/interpret-core/interpret/glassbox
Blackbox and Greybox
- Blackbox Explainers
python/interpret-core/interpret/blackbox - Greybox Explainers
python/interpret-core/interpret/greybox
Model Computation
- Data Preparation
python/interpret-core/interpret/utils/_clean_x.py - Native EBM Engine
shared/libebm
Explanation and Inspection
- Explanation Objects
python/interpret-core/interpret/utils/_explanation.py - Visualization Provider
python/interpret-core/interpret/provider/_visualize.py - Interactive Views
python/interpret-core/interpret/visual/_interactive.py - Comparison Dashboard
python/interpret-core/interpret/visual/dashboard.py - Performance Measures
python/interpret-core/interpret/perf
Privacy and Extensions
- Differentially Private EBM
python/interpret-core/interpret/privacy/_dpebm.py - Extension API
python/interpret-core/interpret/ext/extension.py
Other components
- Data Scientist
- Training Data
- Blackbox Predictor
- Plotting Libraries
Connections 23
- Data Scientist → Training Data: provides
- Training Data → Data Preparation: prepares
- Data Preparation → Explainable Boosting: feeds training
- Explainable Boosting → Native EBM Engine: uses computation
- Explainable Boosting → Explanation Objects: explains globally or locally
- Data Scientist → Other Glassbox Models: fits model
- Other Glassbox Models → Explanation Objects: generates explanation
- Data Scientist → Blackbox Explainers: requests explanation
- Blackbox Predictor → Blackbox Explainers: provides predictions
- Blackbox Explainers → Explanation Objects: produces explanation
- Data Scientist → Greybox Explainers: requests explanation
- Blackbox Predictor → Greybox Explainers: provides model behavior
- Greybox Explainers → Explanation Objects: produces explanation
- Data Scientist → Differentially Private EBM: fits private model
- Differentially Private EBM → Explanation Objects: explains model
- Data Scientist → Visualization Provider: displays explanation
- Explanation Objects → Visualization Provider: supplies view data
- Visualization Provider → Interactive Views: dispatches visualization
- Interactive Views → Plotting Libraries: renders plots
- Data Scientist → Comparison Dashboard: compares explanations
- Explanation Objects → Comparison Dashboard: supplies comparisons
- Data Scientist → Performance Measures: evaluates model
- Data Scientist → Extension API: loads extensions