Figuring out what computers and AI can and cannot do, how they are trained on our data (including personal data!) and how this will affect us, is a pressing challenge. We need to understand the fundamental limitations and capabilities of machine learning, develop principled approaches to data privacy, and make sure nobody is left behind when it comes to accessing these new technologies.
Dr Canonne combines techniques from mathematics and computer science to develop ultra-fast algorithms for machine learning with rigorous guarantees: algorithms which can be run on personal data without revealing any sensitive information, algorithms which can be executed on very small and limited devices, or which can process massive amounts of data very quickly. He designs these algorithms and mathematically analyses their performance, so that we can be certain of what they do, and that we can trust their output.