Research Interests

My main research interests are in the development and application of machine learning methods. Particularly kernel-based learning algorithms, and developing new kernels for structured domains. Specifically new kernels for text, bio-informatics applications, molecular structures; using probabilistic models in these algorithms, and relationship to general graphical models, fisher kernels and probability kernels. I'm also interested in the M-cubed (Max Margin Markov Network) style algorithms in the setting of learning in a structured domain, and finding pre-images from (structured) kernel space. These methods are part of the focus of a new EU STreP project called SMART. I've also recently developed an interest in computer security and applying machine learning methods for intrusion detection and other related tasks. Other interests include strategic game playing, particularly machine learning methods for the games of Poker and Go.

Projects

Current:

• Instant Knowledge MobileVCE, TSB
• SMART (Statistical Multi-lingual Analysis for Retrieval and Translation). EU STreP project
• Kernels for Molecular Structures GlaxoSmithKline
• PASCAL EU Network of Excellence : Pattern Analysis, Statistical Modelling and Computational Learning

Old:

• Development of String-type kernels
• LAVA (Learning for Adaptive Visual Agents)
• KERMIT (Kernel Methods for Images and Text)

Research Activities

At the moment:

• Programme Committee: IJCNN 2006, ESANN 2006, AISTATS 2007
• Reviewing Duties : JMLR, TPAMI, NIPS

In the past I've also run several workshops on kernel methods and SVMs at the NIPS and IJCAI conferences, and at UK NCAF meetings among others, and reviewed papers for a variety of other acronyms. Whilst at Royal Holloway I was part of the coding effort for one of the first publically available SVM implementations, and am currently trying to put together some structured kernel software tools and data. More and more I believe research code (+ data!) should be available on the web in a form so that any experiments are reproducible. Researchers please do this in future! (this includes me!).

Research Team Past and Present

• Research Assistant: Sandor Szedmak (Optimisation methods for Learning with Structured Outputs)
• Research Assistant: Martin Helmhout (Agent learning, Instant Knowledge project
• Research Assistant: Alexei Vinokourov (String kernels/Bioinformatics)
• PhD Student: Tony Demco (Graph Kernels/Chemoinforamtics)
• PhD Student: Bassam Farran (Machine Learning and Security)
• PhD Student: Yizhao Ni (Max-margin methods for Machine Translation)
• PhD Student: David Hardoon (Semantic Kernel Models)

Teaching

Currently I teach the COMP2007 course on Software Analysis and Design, COMP2036 Intelligent Algorithms, COMP2011 Theory of Computing and COMP6021 Adaptive Modelling of Complex data. In the past I've taught courses on Machine Learning, C++, Java, Software Engineering, Software Metrics, Business information systems and Artificial Intelligence.

I've supervised several undergrad projects on topics including Web spiders, Go, Reinforcement learning, Image clustering, Gesture recognition and Spam detection.