The Group's research is mainly, though not entirely, devoted to Inductive Logic Programming: a form of machine learning in which the learning program constructs general rules from specific examples, together with extra "background knowledge". Examples, rules and background knowledge are expressed in a logic programming language such as Prolog.
We have developed various implementations, of which the most recent is Progol, a Prolog interpreter augmented with an inverse entailment algorithm which constructs new clauses by generalising from nominated examples in the Prolog database. Results from the theory of ILP guarantee that Progol will conduct an admissible search through the space of generalisations, finding the maximally compressive set of clauses from which all the examples can be inferred.
The Group's theoretical work extends beyond such results as the above to various topics in the general theory of machine learning, including U-learnability. We are also researching into applications of ILP, including drug design, protein shape prediction and satellite fault diagnosis.
On the non-ILP front, we are using decision-tree techniques to learn rules for selection of the best embryos for transfer in In Vitro fertilisation. Another project involves the comparison of decision-tree methods with ILP in "reverse-engineering" human skills: designing controllers which emulate human F-16 pilots.
We are involved in a joint Anglo-Japanese research project under the auspices of the Machine Intelligence series. Starting in 1995, the series will continue as the proceedings of Anglo-Japanese workshops on Machine intelligence tools for discovery in science and technology, with emphasis on inductive logic programming, neural networks and related statistical learning techniques, and database mining. These will be applied particularly to problems in molecular biology and in dynamic control.
Finally, we maintain a collection of datasets containing example data arising from our work on applications of ILP. They are provided as a service to machine learning researchers wishing to compare and evaluate different machine learning techniques.
The publications of the group can be found on a separate page.
These can be found on a separate page.
A list of forthcoming events can be found on a separate page.
Details of ILP conference proceedings dating back to ILP91 can be found on a separate page.