讲座简介:
|
For one-dimensional continuous time series, the stochastic volatility model and its inference methods have long been a subject of active research. Their extensions to multi-dimensional cases, however, have met a lot challenges ranging from modeling to computation. We here introduce a generalized mixture model and variational Bayes inference procedures for handling multi-response and multidimensional Gaussian processes. More precisely, we propose a mixture model for covariate (time)-dependent covariance matrix of a heteroscedastic Gaussian process, which is computationally more friendly than available approaches. We demonstrate how to use variational approximations to carry out an explicit marginalization of the hidden functions, resulting in efficient parameter estimation and process forecasting. We demonstrate its advantages by both simulations and applications to real-data examples of regression, classification and state-space models. |