Implementation:Interpretml Interpret Merge Ebms

Field	Value
Sources	Repo: InterpretML
Domains	Federated_Learning, Model_Ensembling
Updated	2026-02-07

Overview

Concrete tool for merging multiple EBM models into one provided by the InterpretML library.

Description

The merge_ebms function takes a list of fitted EBM models and produces a single merged model. It validates compatibility, harmonizes bin definitions (union of cut points, merged category dictionaries), remaps score tensors via interpolation, averages across models, and applies process_terms for final aggregation.

Usage

Call this when combining independently trained EBMs for federated learning or ensemble scenarios.

Code Reference

Field	Value
Source	interpretml/interpret
File	python/interpret-core/interpret/glassbox/_ebm/_merge_ebms.py
Lines	280-775

Signature:

def merge_ebms(models):
    """Merges EBM models trained on similar datasets that have the same set of features.
    Args:
        models: List of EBM models to be merged.
    Returns:
        An EBM model with averaged mean and standard deviation of input models.
    """

Import:

from interpret.glassbox import merge_ebms

I/O Contract

Inputs:

Parameter	Type	Required	Description
models	list of fitted EBMs	Yes	Must have the same features and link function

Outputs:

Type	Description
EBM model	A single merged EBM model with averaged scores and combined standard deviations

Usage Examples

from interpret.glassbox import ExplainableBoostingClassifier, merge_ebms

# Train on different data partitions
ebm1 = ExplainableBoostingClassifier()
ebm1.fit(X_train_1, y_train_1)

ebm2 = ExplainableBoostingClassifier()
ebm2.fit(X_train_2, y_train_2)

# Merge into single model
merged = merge_ebms([ebm1, ebm2])
predictions = merged.predict(X_test)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment