Timezone: »
We introduce a scalable approach to Gaussian process inference that combines spatio-temporal filtering with natural gradient variational inference, resulting in a non-conjugate GP method for multivariate data that scales linearly with respect to time. Our natural gradient approach enables application of parallel filtering and smoothing, further reducing the temporal span complexity to be logarithmic in the number of time steps. We derive a sparse approximation that constructs a state-space model over a reduced set of spatial inducing points, and show that for separable Markov kernels the full and sparse cases exactly recover the standard variational GP, whilst exhibiting favourable computational properties. To further improve the spatial scaling we propose a mean-field assumption of independence between spatial locations which, when coupled with sparsity and parallelisation, leads to an efficient and accurate method for large spatio-temporal problems.
Author Information
Oliver Hamelijnck (The Alan Turing Institute)
William Wilkinson (Aalto University)
Niki Loppi (NVIDIA)
Arno Solin (Aalto University)
Theodoros Damoulas (University of Warwick)
More from the Same Authors
-
2021 : Robust Bayesian Inference for Simulator-based Models via the MMD Posterior Bootstrap »
Harita Dellaporta · Jeremias Knoblauch · Theodoros Damoulas · Francois-Xavier Briol -
2022 : Fantasizing with Dual GPs in Bayesian Optimization and Active Learning »
Paul Chang · Prakhar Verma · ST John · Victor Picheny · Henry Moss · Arno Solin -
2022 : Towards Improved Learning in Gaussian Processes: The Best of Two Worlds »
Rui Li · ST John · Arno Solin -
2021 : Sparse Gaussian Processes for Stochastic Differential Equations »
Prakhar Verma · Vincent ADAM · Arno Solin -
2021 Poster: Dual Parameterization of Sparse Variational Gaussian Processes »
Vincent ADAM · Paul Chang · Mohammad Emtiyaz Khan · Arno Solin -
2021 Poster: Dynamic Causal Bayesian Optimization »
Virginia Aglietti · Neil Dhir · Javier González · Theodoros Damoulas -
2021 Poster: Periodic Activation Functions Induce Stationarity »
Lassi Meronen · Martin Trapp · Arno Solin -
2021 Poster: Higher Order Kernel Mean Embeddings to Capture Filtrations of Stochastic Processes »
Cristopher Salvi · Maud Lemercier · Chong Liu · Blanka Horvath · Theodoros Damoulas · Terry Lyons -
2021 Poster: Scalable Inference in SDEs by Direct Matching of the Fokker–Planck–Kolmogorov Equation »
Arno Solin · Ella Tamir · Prakhar Verma -
2020 Poster: Stationary Activations for Uncertainty Calibration in Deep Learning »
Lassi Meronen · Christabella Irwanto · Arno Solin -
2020 Poster: Deep Automodulators »
Ari Heljakka · Yuxin Hou · Juho Kannala · Arno Solin -
2019 Poster: Multi-resolution Multi-task Gaussian Processes »
Oliver Hamelijnck · Theodoros Damoulas · Kangrui Wang · Mark Girolami -
2018 Poster: Infinite-Horizon Gaussian Processes »
Arno Solin · James Hensman · Richard Turner