Advanced Topics in Machine Learning (ATML)

As machine learning models are often trained on data containing sensitive private information, it is important to provide reasonable privacy protection to the individuals whose data is included. This course focuses on advanced topics in privacy-preserving machine learning, with a focus on techniques that maintain data privacy without compromising the performance of the learning algorithm. Key concepts covered include differential privacy and federated learning. By the end, students will be prepared to pursue a master's thesis on privacy in machine learning.

The course is relevant for computer science students as well as students from other studies with a good mathematical background, including Statistics, Actuarial Mathematics, Mathematics-Economics, Physics, etc. We assume that the students have previously passed Machine Learning A+B courses offered by DIKU.

Topics covered include the following broad subjects

• Privacy attacks on machine learning models
• Defining Differential privacy and implementing differentially private algorithms, including differentially private Machine Learning algorithms.
• Learning-theoretic analysis of differentially private machine learning
• Design and Analysis of Federated learning algorithms

WARNING: If you have not taken DIKU's Machine Learning A+B courses, please, check the "Recommended Academic Qualifications" box below. Machine Learning courses given at other places do not necessarily prepare you well for this course. It is not advised taking the course if you do not meet the academic qualifications.