We explore the intersection of deep learning and probabilistic machine learning, addressing limitations of Gaussian processes in comparison to neural networks and proposing …
Our paper "Spherical Inducing Features for Orthogonally-Decoupled Gaussian Processes" was accepted to ICML2023 as an Oral Presentation!
We introduce spherical inter-domain inducing features that yield more flexible, data-dependent basis functions for orthogonally-decoupled GP approximations, narrowing the …
Our paper "Variational Inference for Graph Convolutional Networks in the Absence of Graph Data and Adversarial Settings" was accepted to NeurIPS 2020 as a Spotlight Presentation …
We propose a joint probabilistic model with stochastic variational inference to improve the performance and robustness of graph convolutional networks (GCNs) in scenarios without …
We summarize the notation, identities, and derivations underlying the sparse variational Gaussian process (SVGP) framework.
Amazon Machine Learning Community Tech Talk, Berlin.
We show how to approximate the KL divergence (in fact, any f-divergence) between implicit distributions using density ratio estimation by probabilistic classification.
ICML 2018 Workshop on Theoretical Foundations and Applications of Deep Generative Models (TAGDM), Stockholm.