MAPIE v1 Release Notes

Note

These release notes are kept up to date with the latest version of MAPIE v1.

Introduction

MAPIE v1's primary goal is to make the library easier to use, especially for users unfamiliar with conformal predictions. This release consists of an API and documentation revamp, enhancing MAPIE's clarity and accessibility.

Moreover, the new API structure paves the way for efficient development and internal refactoring in the future.

---

API Changes Overview

The classification and regression APIs have been thoroughly revamped (except for time series). Other API changes (calibration, multi-label classification, time series, etc.) consist mostly of renaming to bring overall consistency and clarity.

High-Level Class Mapping

v0.x class	Corresponding v1 class(es)
MapieRegressor	SplitConformalRegressor, CrossConformalRegressor, JackknifeAfterBootstrapRegressor
MapieClassifier	SplitConformalClassifier, CrossConformalClassifier
MapieQuantileRegressor	ConformalizedQuantileRegressor
MapieTimeSeriesRegressor	TimeSeriesRegressor
MapieMultiLabelClassifier	PrecisionRecallController
MapieCalibrator	TopLabelCalibrator

Key Changes

• The v1 .fit method has been split into .fit and .conformalize for split conformal techniques, and replaced by .fit_conformalize for cross-conformal techniques.
• The alpha parameter has been replaced with confidence_level (confidence_level = 1 - alpha).

Functional Regressions

• The MondrianCP class is temporarily unavailable in MAPIE v1. Mondrian can still easily be implemented manually (tutorial provided).
• In regression settings, the naive method no longer exists for cross conformal techniques.

---

Python, scikit-learn and NumPy support

Dependency	Required version
Python	≥ 3.9
NumPy	≥ 1.23
scikit-learn	≥ 1.4

---

API Changes in Detail

Regression and Classification

Classes

MAPIE v1 breaks down MapieRegressor and MapieClassifier into 5 classes, each dedicated to a particular conformal prediction technique.

The cv parameter is key to understand what new class to use:

Regression:

v0.x cv parameter	Corresponding v1 class	Type
split	SplitConformalRegressor(prefit=False)	Split
prefit	SplitConformalRegressor(prefit=True)	Split
None, integer, or BaseCrossValidator	CrossConformalRegressor	Cross
subsample.Subsample	JackknifeAfterBootstrapRegressor	Cross

Classification:

v0.x cv parameter	Corresponding v1 class	Type
split	SplitConformalClassifier(prefit=False)	Split
prefit	SplitConformalClassifier(prefit=True)	Split
None, integer, or BaseCrossValidator	CrossConformalClassifier	Cross

Workflow

The conformal prediction workflow has been changed for better clarity and control:

v0.x Workflowv1 Workflow

1. Data splitting + model training + calibration → .fit(X, y)
2. Prediction → .predict(X_test, y_test)

1. Data splitting → train_conformalize_test_split()
2. Model training → .fit(X_train, y_train)
3. Conformalization → .conformalize(X_conf, y_conf)
4. Prediction → .predict_interval() / .predict_set()

Conformalization

The calibration step has been named conformalization, to avoid confusion with probability calibration.

For cross conformal techniques, steps 2 and 3 are performed simultaneously using .fit_conformalize().

Parameters

alpha → confidence_level

Replaced with confidence_level (equivalent to 1 - alpha). Now set at initialization.

cv

• v0.x: Used to indicate pretraining or cross-validation scheme.
• v1: Now only relevant to cross conformal techniques. For split conformal, use the new prefit parameter.

conformity_score

• v0.x: Only allowed subclass instances like AbsoluteConformityScore().
• v1: Now also accepts strings, like "absolute".

method

• v0.x: Used in MapieRegressor, relevant only to cross conformal techniques.
• v1: Used only in CrossConformalRegressor and JackknifeAfterBootstrapRegressor. Value "naive" removed.

fit_params and sample_weight

• v0.x: sample_weight was a keyword argument of .fit().
• v1: All fit parameters are passed in the fit_params dictionary.

predict_params

• v0.x: Passed to both .fit() and .predict().
• v1: Passed only to .fit() / .fit_conformalize(), automatically reused at prediction time.

---

Other API Changes

Time Series

• MapieTimeSeriesRegressor → TimeSeriesRegressor
• Takes confidence_level instead of alpha

Risk Control

• MapieMultiLabelClassifier → PrecisionRecallController

Calibration

• MapieCalibrator → TopLabelCalibrator
• The method parameter (only accepting "top-label") has been removed.

Mondrian

The MondrianCP class is temporarily unavailable in v1. A manual implementation tutorial is available.

Metrics

Metrics are divided into three modules: calibration, classification, and regression.

# v0.x
from mapie.metrics import classification_coverage_score

# v1
from mapie.metrics.classification import classification_coverage_score

Conformity Scores

Import from mapie.conformity_scores.bounds or simply mapie.conformity_scores. The deprecated ConformityScore class is replaced by BaseRegressionScore.