Continually learning and unlearning with privacy

As AI systems becomes increasingly prevalent, they must continually learn from fresh data while avoiding repeated training. Many real-world uses involve sensitive data, requiring data privacy safeguards as well as facilitate unlearning of data to address the “right to be forgotten”. This project addresses how to effectively and continually learn and unlearn data while preserving performance. Over five years, this project will support two PhD students and one postdoc.