Infinite latent process decomposition

摘要

This paper presents infinite latent process decomposition (iLPD), a new microarray analysis method, as an extension of latent process decomposition [1]. Our method assumes an infinite number of latent processes. Further, our new collapsed variational) Bayesian inference improves the inference proposed in [2] in the treatment of Dirichlet hyperparameters. We also give the results of the comparison experiment.