Does Multi-Clause Learning Help in Real-World Applications?

The ILP system Progol is incomplete in not being able to derive a multi-clause hypothesis from an example. However, due to the assumption that a multi-clause hypothesis can be built by sequentially adding single clauses, Progol’s incompleteness does not stop it being applied to real-world applications. This paper uses two real-world applications in systems biology to study whether a complete multi-clause learning method MC-TopLog can make a difference to learning results compared to the single-clause learning method Progol5. The experimental results show that in both applications there exist data sets, in which hypotheses derived by MC-TopLog have higher predictive accuracies, as well as better biological significance than those of Progol5.