Overview
The What-If Tool (WIT) is an open-source visual interface designed to understand and debug machine learning models. It facilitates the exploration of model behavior by allowing users to test hypothetical scenarios, analyze feature importance, and visualize performance across multiple models and subsets of data. WIT supports various ML fairness metrics, providing insights into potential biases. It integrates with platforms like Colaboratory, Jupyter notebooks, Cloud AI Notebooks, TensorBoard, TFMA, and Fairness Indicators. Compatible models include TF Estimators, models served by TF Serving, Cloud AI Platform Models, and models wrapped in Python functions. It handles binary classification, multi-class classification, regression, and supports tabular, image, and text data.