Course Description
An introduction to technologies that support building trustworthy machine learning systems. Topics include, but are not limited to, adversarial machine learning, privacy-preserving machine learning, transparency in machine learning, and fairness of machine learning.
Athena Title
Trustworthy Machine Learning
Prerequisite
CSCI 4260/6260 or permission of department
Semester Course Offered
Not offered on a regular basis.
Grading System
A - F (Traditional)
Course Objectives
The objective of this course is to familiarize students with technologies that support building trustworthy machine learning systems. Students will learn about attacks against machine learning models, defense techniques to mitigate such attacks, as well as interpretation methods that can make machine learning models more transparent. This course is appropriate for students who are interested in trustworthy machine learning and want to investigate key opportunities and challenges emerging in the research of this area.
Topical Outline
I. Machine learning overview II. Adversarial machine learning - Data poisoning attacks - Evasion attacks (adversarial examples) - Defenses against poisoning attacks - Defenses against adversarial examples III. Privacy-preserving machine learning - Data inference attacks - Model inference attacks - Privacy-preserving learning IV. Transparency in machine learning - Interpretability - Interpretable models - Model-agnostic methods - Example-based explanations V. Fairness of machine learning - Bias in machine learning - Algorithmic techniques for fairness
Syllabus