Project on Algorithmic Fairness

A special course in Responsible AI @ DTU

In this project, we perform a Gender Fairness analysis to evaluate the impact of removing proxy variables. The evaluated classifier predicts whether some subjects are prone to re-offend or not in their crimes. The approaches developed evaluate Independence, Separation, and Calibration as Fairness Criteria for the mitigation strategy chosen.

Setup

Clone the repository and create a virtual environment (with Python 3.10). A pre-defined environment running with CUDA 11.6 can be created like:

Create environment

Run the following:

conda create -n fairness_ai python=3.10

Install the dependencies:

pip install -r requirements.txt

PyTorch - CPU

If running on CPU install Pytorch with the following command:

pip3 install torch torchvision torchaudio

PyTorch - GPU (CUDA 11.6)

If running on GPU with CUDA 11.6 install Pytorch with the following command:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

Project Organization

---

├── README.md          <- The top-level README for developers using this project.
├── data
│   └── catalan_data_course            <- The original, immutable data dump.
│
├── setup.py           <- makes project pip installable (pip install -e .) so src can be imported
├── src                <- Source code for use in this project.
│   ├── __init__.py    <- Makes src a Python module
│   │
│   ├── data           <- Scripts to download or generate data
│   |   ├── __init__.py
│   │   └── dataloader.py
│   │
│   └── models         <- Scripts to train models and then use trained models to make
│       │                 predictions
│       ├── __init__.py
│       ├── model.py
│       └── train_model.py
│
└── requirements.txt