Home > Categories > Ethical AI and Safety > Transparency and Explainability > Google What-If Tool

Google What-If Tool

Rating:

4.5 / 5.0

Tags

AI, machine learning, model interpretability, fairness, visualization, TensorFlow, XGBoost, Scikit-learn

Categories:
Business AnalyticsEthical AI and SafetySoftware Development
Creator Google
Date 2019-07-18
Platforms Web, Jupyter, Colab, TensorBoard
Status Active
Website pair-code.github.io
Price Model Free (Open Source)
Sections:
Bias AuditingData VisualizationDebuggingTransparency and Explainability

Pricing Details

Free and open-source tool available under the Apache 2.0 license.

Features

Interactive model analysis, Visual exploration of machine learning models, Ability to adjust features and see predictions, Comparison of different scenarios, Easy-to-use interface for non-experts, Supports TensorFlow models, Helps in model interpretation and bias detection.

Integrations

Integrates with TensorBoard, Jupyter, Colab, and supports models from TensorFlow, XGBoost, and Scikit-learn.

Preview

The What-If Tool provides a user-friendly interface to probe the behavior of trained machine learning models. Users can edit data points, explore model predictions, and assess fairness metrics across different subgroups. It integrates seamlessly with TensorBoard, Jupyter, and Colab notebooks, supporting models built with TensorFlow, XGBoost, and Scikit-learn. Key features include counterfactual analysis, partial dependence plots, and performance evaluation across various data slices.