G.: Informative priors for markov blanket discovery

We present a novel interpretation of informa-tion theoretic feature selection as optimiza-tion of a discriminative model. We show that this formulation coincides with a group of mutual information based filter heuristics in the literature, and show how our proba-bilistic framework gives a well-founded ex-tension for informative priors. We then de-rive a particular sparsity prior that recovers the well-known IAMB algorithm (Tsamardi-nos & Aliferis, 2003) and extend it to create a novel algorithm, IAMB-IP, that includes do-main knowledge priors. In empirical evalua-tions, we find the new algorithm to improve Markov Blanket recovery even when a mis-specified prior was used, in which half the prior knowledge was incorrect.