B1127
Title: Spatially-coupled hidden Markov models
Authors: Vianey Leos Barajas - University of Toronto (Canada) [presenting]
Abstract: Hidden Markov models (HMMs) provide a flexible framework to model time series data where the observation process $Y$ is taken to be driven by an underlying latent state process $Z$. HMMs can accommodate multivariate processes by (I) assuming that a single state governs the $M$ observations at time $t$, (ii) assuming that each observation process is governed by its own HMM, or (iii) a balance between the two, as in the coupled HMM framework. Coupled HMMs assume that a collection of $M$ observation processes are governed by their respective $M$ state processes, where the state process for process m at time $t$, depends on all other state processes at time $t-1$. We introduce spatially-coupled hidden Markov models where the state processes interact according to an imposed neighborhood structure with observations collected across $N$ spatial locations. We outline an application to short-term forecasting of wind speed using data collected across meteorological stations.
