After completing honours in pure mathematics at UNSW I went on to do a PhD in machine learning, also at UNSW. During my PhD I interned at IBM Research in New York and worked as a programmer at Proxima Technology and as a machine learning research engineer at Canon Research, both in Sydney.
My research in machine learning ranges from the theory of statistical and sequential prediction (a.k.a. "online learning") and market mechanisms for learning, to developing protocols for presenting learning algorithms as web services. My aim is to build foundations for wide-spread, machine-assisted prediction.
My research in machine learning ranges from the theory of statistical and sequential prediction and market mechanisms for learning to developing protocols for presenting learning algorithms as web services. My aim is to build foundations for wide-spread, machine-assisted prediction.
Machine learning is the study of programs that can discover and exploit patterns in data.
I am especially interested in the theoretical limits of what programs are capable of learning, how quickly they can do it, and what trade-offs are involved. Recently, I have been looking at how the structure of the rewards or penalties a program receives for its predictions affects the rate at which learning can occur.
I am also interested in theory and techniques for composing predictions, such as boosting, prediction markets, and transfer learning. The Protocols and Structures for Inference project that I lead aims to build the technological foundations for exploring these ideas through machine learning algorithms that are presented as web services.
I currently teach into courses on Information Theory (COMP2610) and Convex Optimisation (COMP8650).