This paper presents results comparing three inductive learning systems using different representations for concepts, namely: CNF formulae, DNF formulae, and decision trees. The CNF learner performs surprisingly well. Results on five natural data sets show that it frequently trains faster and produces more accurate and simpler concepts. The probable explanation for its superior performance is that the other systems are more susceptible to the replication problem.

PDF, PS

Machine Learning, Vol. 19, 1 (1995), pp. 79-92.

@Article{mooney:mlj95, title={Encouraging Experimental Results on Learning CNF}, author={Raymond J. Mooney}, volume={19}, journal={Machine Learning}, number={1}, key={rule induction}, pages={79-92}, url="http://www.cs.utexas.edu/users/ai-lab?mooney:mlj95", year={1995} }