A Generalization of Forward-Backward Algorithm

Structured prediction has become very important in recent years. A simple but notable class of structured prediction is one for sequences, so-called sequential labeling. For sequential labeling, it is often required to take a summation over all the possible output sequences, when estimating the parameters of a probabilistic model for instance. We cannot make the direct calculation of such a summation from its definition in practice. Although the ordinary forward-backward algorithm provides an efficient way to do it, it is applicable to limited types of summations. In this paper, we propose a generalization of the forwardbackward algorithm, by which we can calculate much broader types of summations than the existing forward-backward algorithms. We show that this generalization subsumes some existing calculations required in past studies, and we also discuss further possibilities of this generalization.

• Type: Conference
• Year: 2009
• Where: PKDD
• Authors: Ai Azuma, Yuji Matsumoto
• Keywords: Data Mining, Forward-backward Algorithms, PKDD 2009, Sequential Labeling, Structured Prediction