Objective of the sessions

1. Data Anonymization

   • Code Satisfiability of Anonymity Criteria (K-Anonymity, L-Diversity, T-Closeness).

   • Implement Perturbative & Non-Perturbative Methods for satisfying Anonymity Criteria.

   • Understand Trade-Offs in Anonymization and Limitations of different Anonymity Criteria. 

2. Algorithmic Fairness I: Regression

   • Understand Bias Propagation in Regression as a consequence of Omitting Sensitive Variables.

   • Implement Fairness Criteria for Regression (Counterfactual Fairness & Demographic Parity).

   • Work through Case Studies to adapt Fairness Criteria to real-world scenarios. 

3. Algorithmic Fairness II: Classification

   • Recap of Confusion Matrices, Receiver Operating Characteristic and Error Detection Curves.

   • Implement different Pre/In/Post-Processing methods using AIF360 to improve Model Fairness.

   • Work on a Project, from Discrimination-Discovery to Mitigation of Disparate Impact.

Public repository

The notebooks used in the sessions are available under a Creative Commons License on Github at:

https://github.com/Ashwin-19/UPF-FATE-2025