Covariance estimation of spatio-temporal random variables with kronecker product based models

Abstract

Covariance estimation is a widely studied topic. Due to the nature of many prob lems, high-dimensional Spatio-temporal scenarios are considered frequently. In some cases, the true covariance matrices are also expected to have a Kronecker product-based representation. Especially in wind speed analysis, the true covariance matrix can be as sumed to have spatial and temporal Kronecker factors. This thesis studies Kronecker product-based covariance estimation models. Also, a new Kronecker product-based method is proposed with experimentation results showing its performance.