Home

My Group

Publications

Teaching

Software

Interesting Things

Software

Here are some Matlab tools I wrote - if you use them, please cite me.

FEAST. [download]. This is a feature selection toolbox, providing implementations of common mutual information based filter feature selection algorithms. Click the link for more information. MITOOLBOX. [download]. This is a toolbox written by my student Adam Pocock. It provides fast implementations of common mutual information functions such as entropy and conditional mutual information. This was originally implemented to support our research on feature selection - as such there are some demo files bundled with this implementing common information theoretic feature selection algorithms. These demos are early prototypes of the FEAST toolbox. If you use this, please cite the following paper: Conditional Likelihood Maximisation: A Unifying Framework for Information Theoretic Feature Selection G.Brown, A.Pocock, M.Lujan, M.-J.Zhao Journal of Machine Learning Research, vol 13, pages 27-66 (2012)

BNTBASIC. [download]. Kevin Murphy wrote the Bayes Net Toolbox, an incredibly powerful and flexible toolbox for probabilistic modelling. However, sometimes it can be a bit of a "sledgehammer to crack a nut". The BNTBASIC toolbox is designed as an add-on to BNT (i.e. it relies on BNT being installed first). It provides the ability to make nets that are commonly used as classifiers, like Naive Bayes. It supports discrete nodes only, learnt by maximum likelihood. One advantage of these simplifications is that the inference implemented here is approximately 20-25x faster that the BNT toolbox inference engines. Example to train/test a Naive Bayes classifier:      model = bayesnettrain( data, labels, 'naive' );      error = bayesnettest( data, labels, model ); Example to build an arbitrarily connected bayes net classifier:      dag = [ insert my adjacency matrix here ];      model = bayesnettrain( data, labels, dag ); Please cite as:    G.Brown, BNTBASIC Matlab Toolbox, University of Manchester, 2010.    URL: http://www.cs.man.ac.uk/~gbrown/software.php