I find our ability to make inferences about the world fascinating. We are able to extract the general from the particular and reason about the particular using the general.

The small amount of research I’ve done to date has been in machine learning. That is, investigating how these processes might be automated and implemented on computers. More specifically, I’m interested in Bayesian modes of inference, the transfer of inductive bias, and rule learning. Recently, I’ve starting learning about more statistical techniques such as kernel methods and applications of spectral graph theory to dimensionality reduction, classification and regression.

A complete list of my publications can be found at my publications page. Below are the two major pieces of work

My Ph.D. research investigated learning sets of rules from a limited number of training examples. I proposed a novel approach call “similarity-based transfer” that is able to extract and transfer information from one learning task and use it in another.

An implementation of this approach, called DEFT - Description-based Evaluation Function Transfer, was built on top of the ILP system Aleph. The implementation can be found here along with more information and some datasets.

My thesis was submitted on the 14th of July 2006 and was accepted with minor corrections on the 1st of May 2007.

DEFT Guessing: Using Inductive Transfer to Improve Rule Evaluation from Limited Data (2.3Mb PDF)

My undergraduate degree was a Bachelor of Science majoring in pure mathematics and computer science. For my honours year project I looked at an open problem in harmonic analysis. The problem involves determining the arrangement of radial slits on a unit disc that maximises the value of a harmonic function at the centre of the disc when the function takes the value 1 on the slits and zero on the disc’s circumference.

The dissertation I produced for that work is unsurprisingly entitled “Harmonic Measure on Radially Slit Disks”. In it I look at two proofs of special cases of the general problem, provide some computation “evidence” for more general cases an propose a conjecture for slits of uneven lengths.

Harmonic Measure on Radially Slit Disks (1Mb PDF)