The relational concept learner HYDRA extends
the machine learning program FOCL by adding likelihood ratios to the induced
classification rules. HYDRA learns a concept description for each
class. The concept descriptions compete to classify test examples based on the
likelihood ratios that are assigned to clauses of that concept description.
This makes the algorithm more robust against noise.
References
K.M. Ali and M.J. Pazzani.
Hydra: A noise-tolerant relational concept learning algorithm.
In R. Bajcsy, editor, Proceedings of the 13th International
Joint Conference on Artificial Intelligence, pages 1064-1071. Morgan
Kaufmann, 1993.