Sérgio Jesus

Separated setups

1

Fixes #108 `setup.py` supports now a new argument `-l` (or `-lite`). Requirements are separated, and only the required files are copied when creating a wheel. This MR changes the way...

Projects that use `aequitas` might not require the `cli` package and the `webapp` package. Because of this, making a setup where only the `aequitas` package is installed might be of...

Add two new columns to the result of the `Group.get_crosstabs` method, representing Group-Size adjusted False Positives and Group-Size adjusted False Negatives respectively. Additionally, make disparities of the two columns available...

Including: - Getting started (install and run) - Minimal examples (load BAF and train a model in a 4-5 lines code example) - Detail FairGBM-specific parameters - Detail inner workings...

Move included datasets to other place

1

Currently the datasets are being hosted in Github. While it works for now, it is not the best practice. Study the best place to leave the dataset (huggingface? OpenML?) and...

Update GitHub pages

1

The current solution for documentation generation for GitHub pages ([link here](https://dssg.github.io/aequitas/index.html#)) has some flaws. These are: 1. New packages, classes, etc. are not included and the process to include them...

@reluzita

Add a readme or page on the existing metrics of fairness

2

We can change the landing page readme, or better yet create a readme that will be used in the sphinx pages. The current one is not great in terms of...

Check `typeguard` package, to see if it is a good option for most of the code.