Models

This section explains how to obtain a model compatible with PyXAI to compute explanations. There are three ways to do so:

• Build a model by hand, using tree nodes and leaves via the Builder module. This is useful for testing and verifying explanations on controlled examples.
• Train a model with PyXAI’s Learning module, which wraps Scikit-learn, XGBoost and LightGBM and handles the full experimental protocol (train/test split, cross-validation, metrics).
• Import an existing model that has already been trained outside of PyXAI, using the ModelIO class.

Once a model is obtained, it can be saved and reloaded at any time using ModelIO.

Table of contents

• Hand Crafting
• Training Models
• Importing Models
• Saving/Loading Models