Gavin Brown

Reader in Machine Learning and Director of Research, School of Computer Science

Gavin Brown

For me, when a student develops a confidence in their own abilities is a good measure of success. I hope we instil the ability to learn new skills, and a confidence going out into the real world with them. Knowing what you know, thus knowing where you would best fit in a job role, is a very important skill.

I have worked at The University of Manchester for nine years, and one of the biggest things that keeps me is the opportunities that come up here. I have worked with world-famous people, and I get lots of support from fellow staff. Ultimately you get out whatever you put into the system – if you contribute a lot, it’s recognized, and you get a lot out of being here in return.

The aspects of our School that make us distinctive are the size of the staff body, and the efficiency of the administration structures, which combine to mean there are several layers of support mechanisms for students. From our triple- and quadruple-checking of exam results, through to pastoral care and helping to organise social events, I do not know of another institution with this level of student support.

My research

The research I work on is in Machine Learning; a field that creates mathematical models which can make predictions about phenomena given some background data, such as predicting whether someone is at risk of heart disease given their medical history, or predicting the price of a car given its characteristics.

The specific area I work on is called 'Feature Selection'. To understand this, imagine you're trying to predict the price of a car, and you are provided with various pieces of information; like the make and model, year of manufacture, etc. In solving any given problem like this, some pieces of information are relevant, some are irrelevant, and some are redundant in the context of others.

In the car example, the number of miles on the clock clearly matters, while the colour of the wheel trim probably does not. The age of the car is probably redundant if you know the mileage. You know this because you (probably) know something about cars.

What about predicting whether someone will have a relapse of a particular cancer? What things matter? Genetic factors? Lifestyle? Metabolic? In my field we use statistics to automatically identify these factors. In particular I work on information theory and other probabilistic methods. Recently, I am particularly interested in incorporating human knowledge into the statistical processes, and dealing with non-standard types of features, such as graph structures.

Teaching

I get a lot of pleasure from seeing students in that ‘ah-ha’ moment when they finally get the idea of something complicated.

For me, when a student develops a confidence in their own abilities is a good measure of success. I hope we instil the ability to learn new skills, and a confidence going out into the real world with them. Knowing what you know, thus knowing where you would best fit in a job role, is a very important skill.

 

Find out more about Gavin and his research.

▲ Up to the top