A simple hierarchical infinite HMM with efficient inference

Dec. 1, 2015


Ardavan Saeedi, Asma Ghandeharioun, Matt Hoffman


We propose a simple hierarchical infinite HMM (iHMM) model, an extension to (iHMM) with efficient inference scheme. The model can capture dynamics of a sequence in two timescales and does not suffer from the problems of other related models in terms of implementation and time complexity. We use the model to analyze the dynamics in two timescales of some synthetic and real physiological data. We show that the model performs reasonably well compared to a baseline on two physiological datasets.

Related Content