Model Distillation and Interpretability Methods for Machine Learning Models
Constructs an ALE for a model.
Build grid used for weights in distilled surrogate model
Centers the predicted values for 1-d ICE and PDP plots or 2-d PDP plot...
Builds surrogate model from an interpreter object based on the univari...
Interpreter class description
Given a interpreter object with at least one pair of features, create ...
Plotting method for Interpretor model
Predict method for Predictor class
Prediction method for the distilled surrogate model
Prediction function for the ALE plots
Prediction Function for ICE Plots
Prediction Function for PDP Plots
Two Dimensional Prediction Curve for PDP Plots
Predictor class description
The Printing method for Predictor class
Sets a new center in the PDP and ICE plots made by an Interpreter
Sets grid points used for plotting PDP and ICE plots
Surrogate class description
Provides several methods for model distillation and interpretability for general black box machine learning models and treatment effect estimation methods. For details on the algorithms implemented, see <https://forestry-labs.github.io/distillML/index.html> Brian Cho, Theo F. Saarinen, Jasjeet S. Sekhon, Simon Walter.
Useful links