Team work on Inductive Logic Programming
The goal of this task is to gain experience with Inductive Logic
Programming (ILP) systems. The work should be done using the Aleph ILP
system. The following topics should be experimented:
- Download an ILP data-set (ex: krki, carcinogenesis).
- Use a Best-First Search (BFS) and either a Randomized Rapid
Restart (RRR) search or a Stochastic Clause Selection (SCS)
search technique and evaluate the impact on accuracy and
performance of such search techniques. Fine tune the most
important parameters for each search technique.
- Use the Theory learning facility to improve the best
theory constructed in the previous step.
We will value:
- the choices made to perform the experiments, and/or,
- the discussion of the results achieved, and/or,
- suggestions to improve the techniques.
Evaluation
The work is to be done by a team of 2 students. Each team must choose
its own set of two search techniques. The report must be sent by email
to Ricardo Rocha until
February 22, 2008. The report should consist of up to 25
pages, written in English and should always include:
- Student identification.
- Report Title.
- An Introduction, stating the experiments done and motivating the
report.
- The description of the search techniques evaluated
- The settings of the experiments
- The results obtained and the discussion of such results
- Conclusion
As discussed above, we will value the report on whether it shows
understanding of and critical opinion of the research. Further,
reports must be concise and clear.
Originality on the discussion of the related work, experiments with
the subject of the report, and being able to connect to one's work
will improve the score.
Information sources
The following papers are good starting points:
- Oxford University repository of ILP applications\\
URL:http://web.comlab.ox.ac.uk/oucl/research/areas/machlearn/applications.html
- York University repository of ILP applications\\
URL:http://www.cs.york.ac.uk/mlg/
- Imperial College repository of ILP applications\\
URL:http://www.doc.ic.ac.uk/~shm/index.html
- Aleph ILP system\\
URL:http://www.comlab.ox.ac.uk/oucl/research/areas/machlearn/Aleph/aleph.pl
- Aleph manual\\
URL: http://web.comlab.ox.ac.uk/oucl/research/areas/machlearn/Aleph/aleph_toc.html
- Yap Prolog site:\\
URL:http://www.ncc.up.pt/~vsc/Yap/
- SWI Prolog site:\\
URL:http://www.swi-prolog.org/
Please do feel free to consider other data-sets!
Good Work!