| Course label : | Advanced Machine Learning 3 - Fairness in Thrustworthy Machine Learning |
|---|---|
| Teaching departement : | EEA / Electrotechnics - Electronics - Control Systems |
| Teaching manager : | Mister PIERRE-ANTOINE THOUVENIN / Mister PIERRE CHAINAIS |
| Education language : | |
| Potential ects : | 0 |
| Results grid : | |
| Code and label (hp) : | MR_DS_S3_AM3 - Advanced machine learning 3 |
Education team
Teachers : Mister PIERRE-ANTOINE THOUVENIN / Mister PIERRE CHAINAIS
External contributors (business, research, secondary education): various temporary teachers
Summary
Nowadays, machine learning methods find widespread application due to their ability to learn models that per-form outstandingly, sometimes reaching human-level capabilities. However, if these models have the potential to impact human lives, for example in justice, solely evaluating them in terms of accuracy is not sufficient any-more. Other notions then need to be considered to ensure that the models are trustworthy and that they do not have a negative impact on individuals. This course will focus on fairness as a trustworthiness concept that aims to eliminate discrimination in machine learning models. First, we will recognize various sources of unfairness before exploring several of the numerous fairness metrics proposed in literature. Then, focusing on a family of measures, we will study several strategies to mitigate these discriminatory behaviors. Finally, we will examine how privacy, another important aspect of trustworthiness, can affect fairness.24 hours, 6 lectures, 6 practical sessions.
Educational goals
The goal of this course is to provide an introduction to the problem of fair machine learning, a core concept for building more trustworthy systems. After successfully completing this course, a student should: • Be able to identify machine learning problems where fairness issues may arise. • Be able to evaluate the degree of unfairness of the models with various metrics. • Be able to deploy some existing solutions to mitigate the discriminatory behaviors. • Be aware of the potential impact that privacy may have on fairness.
Sustainable development goals
Knowledge control procedures
Continuous Assessment
Comments: Practicals, (min) 0 - 20 (max)
Written exam, (min) 0 - 20 (max)
Passing grade 10/20
Online resources
• "Fairness and Machine Learning: Limitations and Opportunities", Solon Barocas, Moritz hardt, Arvind Narayanan • "The Algorithmic Foundations of Differential Privacy", Cynthia Dwork, Aaron Roth
Pedagogy
24 hours, 6 lectures, 6 practical sessions. English is the default language.
Sequencing / learning methods
| Number of hours - Lectures : | 12 |
|---|---|
| Number of hours - Tutorial : | 12 |
| Number of hours - Practical work : | 0 |
| Number of hours - Seminar : | 0 |
| Number of hours - Half-group seminar : | 0 |
| Number of student hours in TEA (Autonomous learning) : | 0 |
| Number of student hours in TNE (Non-supervised activities) : | 0 |
| Number of hours in CB (Fixed exams) : | 0 |
| Number of student hours in PER (Personal work) : | 0 |
| Number of hours - Projects : | 0 |
Prerequisites
• Basic notions of probabilities and statistics • Basic notions of linear algebra • Core concepts in machine learning (e.g. supervised learning, empirical risk minimization, gradient de-scent, …) • Python programming (e.g. jupyter notebook, numpy, pandas, scikit-learn, ...)