CMPUT 622 - Privacy and Fairness in ML

Overview

This course focuses on privacy and fairness in Machine Learning. Privacy breaches and unfairness in ML-based decision-making are important impacts of ML algorithms. This course will take a fundamental look into the definitions of these concepts, how they arise in ML outcomes, and what mitigation methods are state of the art. There will be a significant theory component where the students will be expected to follow the fundamental theory underpinning these concepts.

The objective of the course is to develop research methods on these areas. After some sessions on setting the fundamentals, the course will focus on current research and students will have the opportunity to work on a research project.

Objectives

The course has the following objectives:

• Gain an understanding of the fundamental definitions and sources of privacy and fairness in ML
• Understand state of the art mitigation techniques
• Learn research methods through reviewing and critiquing current methods
• Gain experience through implementing current methods
• Gain research experience through a team project on original research.

Course Topics

• Notions of privacy
• Differential Privacy
• Privacy attacks
• Privacy-preserving ML
• Notions of fairness
• Unfairness mitigation

Course Work

The goal of the course is to gain a fundamental understanding of privacy and fairness and become familiar with the algorithms for mitigation of privacy attacks and unfairness. The students will improve their research skills through the course work which will include seminar-style presentations and a team research project. Active participation is a required component in the course.

• Assignments
• Paper Reviews
• Seminar-style presentations
• Research project