Holistic AI

The Holistic AI library is an open-source tool to assess and improve the trustworthiness of AI systems.

Currently, the library offers a set of techniques to easily measure and mitigate Bias across numerous tasks. In the future, it will be extended to include tools for Efficacy, Robustness, Privacy and Explainability as well. This will allow a holistic assessment of AI systems.

• Documentation: https://holistic-ai.readthedocs.io/en/latest/
• Tutorials: https://github.com/holistic-ai/holisticai/tree/main/tutorials
• Source code: https://github.com/holistic-ai/holisticai/tree/main
• Holistic Ai website: https://holisticai.com

Installation:

For metrics, you can use the default installation:

pip install holisticai # basic installation
pip install holisticai[bias] # bias mitigation support
pip install holisticai[explainability] # for explainability metrics and plots
pip install holisticai[all] # install all packages for bias and explainability

Troubleshooting

on macOS could be necessary some packages before install holisticai library:

brew install cbc pkg-config
python -m pip install cylp
brew install cmake

Explainability Visualization Tools

Install GraphViz

sudo apt update
sudo apt-get install graphviz

Contributing

We welcome contributions to the Holistic AI library. If you are interested in contributing, please refer to our contributing guide.