Mondrian Trees

Mondrian Trees are a type of decision tree built using Mondrian processes, which provide a Bayesian nonparametric approach to tree construction.

Prediction modes

Opinionated default: constant predictions

By default (path_smoothing=False) predictions are piecewise-constant leaf values: each sample follows a single root-to-leaf path and receives the leaf's stored value. This matches scikit-learn's tree and forest behaviour and the plain ONNX ai.onnx.ml tree-ensemble export, but deliberately deviates from the pure Mondrian-process predictor.

Set path_smoothing=True at construction (or pass path_smoothing=True per predict() / predict_proba() call) to restore the original Mondrian-process weighting, where every node on the decision path contributes to the prediction and far-away samples shrink towards the root statistics.

MondrianTreeRegressor

A single Mondrian tree for regression tasks.

from shinrin import MondrianTreeRegressor

tree = MondrianTreeRegressor(
    max_depth=None,
    min_samples_split=2,
    random_state=0,
)
tree.fit(X, y)

Parameters

Parameter	Type	Default	Description
max_depth	int	None	Maximum depth of the tree
min_samples_split	int or float	2	Minimum samples required to split a node
random_state	int	None	Random seed for reproducibility
path_smoothing	bool	False	Use pure Mondrian-process weighted-path predictions instead of constant leaf values

Attributes

Attribute	Type	Description
n_features_in_	int	Number of features seen during fit
tree_	Tree	The underlying tree structure

Methods

Method	Description
fit(X, y)	Fit the Mondrian tree to training data
predict(X)	Predict regression values for X
apply(X)	Return the index of the leaf for each sample
pred_contribs(X)	TreeSHAP values with base value appended (see below)

SHAP values via pred_contribs()

Both Mondrian trees and forests expose pred_contribs(X), which returns TreeSHAP contributions such that prediction = base_value + sum(shap_values) under the estimator's active prediction mode. The last column/axis holds the base value: (n_samples, n_features + 1) for regression, (n_samples, n_features + 1, n_classes) for classification.

contribs = tree.pred_contribs(X)
shap_values, base_value = contribs[:, :-1], contribs[:, -1]

For most purposes the higher-level TreeExplainer / explanation() helpers are more convenient; see TreeSHAP Explanations.

MondrianTreeClassifier

A single Mondrian tree for classification tasks.

from shinrin import MondrianTreeClassifier

tree = MondrianTreeClassifier(
    max_depth=None,
    min_samples_split=2,
    random_state=0,
)
tree.fit(X, y)

Parameters

Same as MondrianTreeRegressor with additional:

Parameter	Type	Default	Description
class_weight	dict	None	Weights associated with classes

Methods

Method	Description
fit(X, y)	Fit the Mondrian tree to training data
predict(X)	Predict class labels for X
predict_proba(X)	Predict class probabilities for X
apply(X)	Return the index of the leaf for each sample