Gizli Degi§kenler if;eren Modellerde Skor E§leme (Score Matching for Models with Latent Variables)
Onur Dikmen, Cnrs Ltci · 2011
Undirected graphical models such as Markov random fields or Boltzmann machines prove useful in many signal processing and machine learning tasks. However, parameter estimation in these models is diffi cult due to the intractable normalising con stant in their probability density functions. One powerful tech nique for parameter estimation in such models is score match ing. This technique makes use of an objective function which is independent of the normalising constant and constitutes locally consistent estimators for the parameters of such models. How ever, score matching is only applicable to fully-observed mod els. In this paper, we extend the applicability of score matching to models with latent variables. Our estimators are unbiased, based on Monte Carlo integration. Unbiased gradient estima tors open the way to optimisation through stochastic approxi mation. We demonstrate the performance of our methodology on two synthetic problems.