Robust/Differentially Private Machine Learning
The topic is flexible and depends on student's background, mathematical knowledge, previous research experience. Generally, this project mainly focuses on how to design robust (especially robust against to outliers or heavy-tailed distributions) or private (or forgettable) algorithms for some foundamental problems in machine learning, deep learning or statistics. Students will provide theoretical guarantees via using mathematical tools from probability, learning theory, optimization and high dimensional statistics. Also, student will analyze utility-privacy tradeoff or robustness-utility tradeoff.
Applied Mathematics and Computer Science
Computer, Electrical and Mathematical Sciences and Engineering
Field of Study -
Machine Learning, Data Privacy, High Dimensional Statistics