DrClement Canonne

Collecting, analysing, and processing large volumes of data, often high-dimensional, has become a centerpiece of modern computer science, machine learning, and industry at large. The time and memory constraints associated with these amounts of information present unprecedented challenges, as classical algorithms and statistical techniques are no longer sufficient to analyse the data. Further, new and even more stringent constraints have emerged, such as critical privacy concerns, or the physical limitations of battery- and communication-limited devices. Dr. Clément Canonne's research focuses on formalising those challenges, and developing sound theoretical foundations to address them.

"My research lies at the intersection of algorithms, information theory, statistics, and computational learning theory, and asks questions of the following flavour. When trying to perform a specific task on very large datasets, do we need exact answers or are approximate ones good enough? Must we store the data itself, or is some concise representation sufficient for our applications? Can interactivity or randomisation help reducing the computational load? What are the tradeoffs, if any, that one can achieve or must incur between privacy, speed, and accuracy?

"Overall, the goal is to analyse the fundamental and practical questions that emerge from the massive amounts of data available. I strive to develop new tools and techniques to analyse the many settings that arise, focusing not only on the purely statistical constraints at play, but also on the new computational and societal aspects that are now at the front and center of data science."