This repository provides the code for reproducing the results obtained in the paper Ensuring Fairness Stability for Disentangling Social Inequality in Access to Education: the FAiRDAS General Framework published at IJCAI24 (AI for Good track).
- Virtual environment with Python 3.7
- Packages:
pip install -r requirements.txt
The script run.py replicates the results presented in the paper:
python run.py
Results are stored in results folder.
In particular, the folder results/records contains:
- the configuration file with all the run information
config.json - the historical records of actions, metrics of interest, optimization and ranking (
.pklfiles) - the mean and standard deviation of the metrics (
statistics.txtandstatistics.json)
while the folder results/images contains images both in png and eps format.
The MLP regressor has been trained on utils/training_data.csv with script utils/train_regressor.py.
The resulting model weights are stored in utils/regressor_checkpoint.pt.
To re-train the regressor, run:
python utils/train_regressor.py
Eleonora Misino: eleonora.misino2@unibo.it