UQ360
Uncertainty Quantification 360 (UQ360) is an extensible open-source toolkit that can help you estimate, communicate and use uncertainty in machine learning mode
What is UQ360?
Uncertainty Quantification 360 (UQ360) is an extensible open-source toolkit that can help you estimate, communicate and use uncertainty in machine learning model predictions.
Overview of UQ360
What it is
Uncertainty Quantification 360 (UQ360) is an open-source toolkit with a Python package that provides data science practitioners and developers access to state-of-the-art algorithms for estimating, evaluating, improving, and communicating uncertainty of machine learning models. The toolkit includes an interactive experience, tutorials, example notebooks, and a complete API. It is designed with extensibility in mind and is still in development.
Key features
- Provides algorithms for uncertainty estimation, metrics, and applications.
- Includes an interactive experience for a gentle introduction to concepts and capabilities.
- Offers tutorials and example notebooks for a deeper, data scientist-oriented introduction.
- Supports meta-models to augment sklearn's gradient boosted regressor with prediction intervals.
- Includes UQ360 metrics for model selection, such as prediction interval coverage probability (PICP) score.
- Compatible with Python 3.7 on macOS, Ubuntu, and Windows.
Use cases
- Augmenting machine learning model predictions with prediction intervals using meta-models.
- Selecting models through cross-validation using UQ360 metrics like PICP.
- Estimating, evaluating, improving, and communicating uncertainty in AI models for transparency.
Getting started
To install UQ360 via pip without cloning the repository:(your environment)$ pip install uq360To install from source after cloning:(uq360)$ git clone https://github.ibm.com/UQ360/UQ360(uq360)$ pip install -e .For detailed setup instructions, including creating a virtual environment with Conda, see the repository README.
License and pricing
UQ360 is licensed under Apache-2.0. It is an open-source toolkit with no stated pricing.
Based on the project's public README and repository metadata.