Module learner 

Nuisance learners: the Regressor / Classifier traits and the built-in implementations.

Structs

GradientBoostingClassifier

Gradient boosting for 0/1 classification (log loss). Defaults match scikit-learn’s GradientBoostingClassifier.

GradientBoostingRegressor

Gradient boosting for regression (squared error). Defaults match scikit-learn’s GradientBoostingRegressor: 100 stages, learning rate 0.1, depth-3 trees.

GridSearchClassifier

Grid search over classifiers, scored by the log loss of the held-out folds (probabilities clipped to [1e-15, 1 − 1e-15]).

GridSearchRegressor

Grid search over regression learners, scored by the mean squared error of the held-out folds.

Lasso

Lasso regression with an unpenalised intercept, fitted by cyclic coordinate descent.

LassoCV

Lasso with the penalty chosen by K-fold cross-validation, as scikit-learn’s LassoCV.

LogisticRegression

L2-penalised logistic regression with an unpenalised intercept, fitted by Newton’s method with backtracking line search.

LogisticRegressionCV

L2-penalised logistic regression with C chosen by stratified K-fold cross-validation, as scikit-learn’s LogisticRegressionCV.

Ols

Ordinary least squares with an intercept and optional ridge penalty.

RandomForestClassifier

Random forest classifier for 0/1 labels; predict_proba averages the class-1 frequency of the leaves.

RandomForestRegressor

Random forest regressor (squared-error CART trees on bootstrap samples).

Enums

CvScoring

Criterion for choosing C in LogisticRegressionCV.

MaxFeatures

Number of features considered at each split.

Traits

Classifier

A classification learner returning P(label = 1 | X), used for propensity scores in models with a binary treatment such as Irm.

Regressor

A regression learner for nuisance functions such as E[Y | X] or E[D | X].