Efficient Learning of Interpretable Classification Rules