FairML

This project addresses the bias present in Machine Learning algorithms. Existing algorithms using preprocessing, in-processing and post-processing have been used. These algorithms are also paired and tested in an emsemble architecture.

This project uses the following algorithms for experimentation:

1. Disparate Impact Remover (pre-processing)
2. Exponentiated Gradient Reduction (in-processing)
3. Equalized odds (post-processing)

The data following datasets have been used for analysis:

1. German dataset
2. Adult dataset