Timezone: »
We present a framework for transfer learning based on modular variational Gaussian processes (GP). We develop a module-based method that having a dictionary of well fitted GPs, each model being characterised by its hyperparameters, pseudo-inputs and their corresponding posterior densities, one could build ensemble GP models without revisiting any data. Our method avoids undesired data centralisation, reduces rising computational costs and allows the transfer of learned uncertainty metrics after training. We exploit the augmentation of high-dimensional integral operators based on the Kullback-Leibler divergence between stochastic processes to introduce an efficient lower bound under all the sparse variational GPs, with different complexity and even likelihood distribution. The method is also valid for multi-output GPs, learning correlations a posteriori between independent modules. Extensive results illustrate the usability of our framework in large-scale and multi-task experiments, also compared with the exact inference methods in the literature.
Author Information
Pablo Moreno-Muñoz (Technical University of Denmark (DTU))
Antonio Artes (Universidad Carlos III de Madrid)
Mauricio Álvarez (University of Sheffield)
More from the Same Authors
-
2021 Poster: Learning Nonparametric Volterra Kernels with Gaussian Processes »
Magnus Ross · Michael T Smith · Mauricio Álvarez -
2021 Poster: Compositional Modeling of Nonlinear Dynamical Systems with ODE-based Random Features »
Thomas McDonald · Mauricio Álvarez -
2020 Poster: Multi-task Causal Learning with Gaussian Processes »
Virginia Aglietti · Theodoros Damoulas · Mauricio Álvarez · Javier González -
2019 Poster: Multi-task Learning for Aggregated Data using Gaussian Processes »
Fariba Yousefi · Michael T Smith · Mauricio Álvarez -
2018 Poster: Heterogeneous Multi-output Gaussian Process Prediction »
Pablo Moreno-Muñoz · Antonio Artés · Mauricio Álvarez -
2018 Spotlight: Heterogeneous Multi-output Gaussian Process Prediction »
Pablo Moreno-Muñoz · Antonio Artés · Mauricio Álvarez -
2017 : Final remarks »
Alessandra Tosi · Alfredo Vellido · Mauricio Álvarez -
2017 : Opening remarks »
Alessandra Tosi · Alfredo Vellido · Mauricio Álvarez -
2017 Workshop: Transparent and interpretable Machine Learning in Safety Critical Environments »
Alessandra Tosi · Alfredo Vellido · Mauricio Álvarez -
2017 Poster: Efficient Modeling of Latent Information in Supervised Learning using Gaussian Processes »
Zhenwen Dai · Mauricio Álvarez · Neil Lawrence