Viterbi training in PRISM
Taisuke Sato, Keiichi Kubota · 2012
VT (Viterbi training), or hard EM, is an efficient way of parameter learning for probabilistic models with hidden variables. Given an observation y, it searches for a state of hidden variables x that maximizes p(x, y | θ) by coordinate ascent on parameters θ. In this paper we introduce VT to PRISM, a logic-based probabilistic modeling system for generative models. VT improves PRISM’s probabilistic modeling in two ways. First although generative models are said to be inappropriate for discrimination tasks in general, when parameters are learned by VT, models often show good discrimination performance. We conducted two parsing experiments with probabilistic grammars while learning parameters by a variety of inference methods, i.e. VT,EM,MAP and VB. The result is that VT achieves the best parsing accuracy among them in both experiments. Second since VT always deals with a single probability of a single explanation, Viterbi explanation, the exclusiveness condition imposed on PRISM programs is no more required when we learn parameters by VT. PRISM with VT thus allows us to write inclusive clause bodies, learn parameters and compute Viterbi explanations. 1.