Mixtures of Gaussian process priors

摘要

Mixtures of Gaussian process priors allow the flexible implementation of complex and situation specific a priori information. This is essential for tasks with, compared to their complexity, small number of available training data. The paperconcentrates on the formalism for Gaussian regression problems where prior mixture models provide a generalisation of classical quadratic, typically smoothness related, regularisation approaches being more flexible without having a much largercomputational complexity.