Decision-Theoretic Logic Programs

We propose a new framework, Decision-theoretic Logic Programs (DTLPs), in the paper. DTLPs extend Probabilistic ILP models by integrating desicion-making features developed in Statistical Decision Theory area. Both decision-theoretic knowledge (eg. utilities) and probabilistic knowledge (eg. probabilities) can be represented and dealt with in DTLPs. An implementation of DTLPs using Stochastic Logic Programs is introduced and a DTLP parameter learning algorithm is discussed accordingly. The representation and methods are tested by performing regression on the traditional mutagenesis dataset.