The Rényi Gaussian Process

In this article we introduce an alternative closed form lower bound on the Gaussian process ($\mathcal{GP}$) likelihood based on the R\'enyi $\alpha$-divergence. This new lower bound can be viewed as a convex combination of the Nystr\"om approximation and the exact $\mathcal{GP}$. The key advantage of this bound, is its capability to control and tune the enforced regularization on the model and thus is a generalization of the traditional sparse variational $\mathcal{GP}$ regression. From the theoretical perspective, we show that with probability at least $1-\delta$, the R\'enyi $\alpha$-divergence between the variational distribution and the true posterior becomes arbitrarily small as the number of data points increase.