Transparentai

v0.2.1 release of the TransparentAI's library.

Change library to match with EU Commission report on Trust AI.

**Difference with previous version :**

- Code has less classes and is more functionnal
- The logic is not to answer "How to create an ethical AI ?" but "Is my AI ethical ?", answering by submodules

Features :

- `datasets` : analyse a structured dataset
- `models` : analyse classification or regression performance
- `models.explainers` : understand global and local feature influence for tree and linear models
- `fairness` : analyse dataset and model bias bias
- `monitoring` : Compare new performances with training
- `utils` : check safety and compute kWh for training

PyPI link : https://pypi.org/project/transparentai/0.2.1/

First release of the TransparentAI's library.

Features :
* `start` : start the ai by knowing if your ai is reliable
* `datasets` : analyse a structured dataset
* `models` : analyse classification or regression performance
* `explainer` : understand global and local feature influence for tree and linear models
* `fairness` : analyse dataset and model bias bias
* `monitoring` : Compare new performances or bias with training

PyPI link : https://pypi.org/project/transparentai/0.1.0/