CFE 2019: Start Registration
View Submission - CMStatistics
B1007
Title: Kernel regression with convolved Gaussian processes on Riemannina manifold Authors:  Jinzhao Liu - Newcastle University (United Kingdom) [presenting]
Abstract: The data analysis on vector spaces is well studied. However, for some new and popular topic, such as computer vision and medical image analysis, the data are often mapped onto a special non-Euclidean space. Most of current models do not work since the data lack of vector structure. With the motivation of solve this problem, we try to find a model which can find the relationship between real-valued covariate and manifold-valued response variable. This model includes two part: one is the mean structure which is a generalisation of kernel method to Riemannian manifold, the other is covariance structure which is based on wrapped Gaussian process on Riemannian manifold. The purpose is to derive a concurrent model for manifold-valued data. In addition, uncertainty and random error are also considered in this mode.l