An l1 Regularization Framework for Optimal Rule Combination

In this paper l1 regularization is introduced into relational learning to produce sparse rule combination. In other words, as few as possible rules are contained in the final rule set. Furthermore, we