Title: Output-sparse latent Gaussian processes
Authors: Markus Heinonen - Aalto University (Finland) [presenting]
Abstract: Zero-inflated datasets, which have an over-abundance of zero outputs, are commonly encountered in problems such as climate or rare event modelling. Conventional machine learning approaches tend to overestimate the non-zeros leading to poor performance. We propose a novel family of zero-inflated Gaussian processes (ZiGP), produced by sparse kernels through learning latent probit processes that can zero out kernel rows and columns whenever the signal is absent. The ZiGPs are particularly useful for making the powerful Gaussian process networks more interpretable. We introduce sparse GP networks where variable-order latent modelling is achieved through sparse mixing of latent signals. To scale the models to large datasets, we demonstrate that the variational inference of probit-sparsified networks of latent Gaussian processes is remarkably tractable.