Timezone: »
Kernel selection plays a central role in determining the performance of Gaussian Process (GP) models, as the chosen kernel determines both the inductive biases and prior support of functions under the GP prior. This work addresses the challenge of constructing custom kernel functions for high-dimensional GP regression models. Drawing inspiration from recent progress in deep learning, we introduce a novel approach named KITT: Kernel Identification Through Transformers. KITT exploits a transformer-based architecture to generate kernel recommendations in under 0.1 seconds, which is several orders of magnitude faster than conventional kernel search algorithms. We train our model using synthetic data generated from priors over a vocabulary of known kernels. By exploiting the nature of the self-attention mechanism, KITT is able to process datasets with inputs of arbitrary dimension. We demonstrate that kernels chosen by KITT yield strong performance over a diverse collection of regression benchmarks.
Author Information
Fergus Simpson (Secondmind)
Ian Davies (InstaDeep)
Vidhi Lalchand (University of Cambridge)
Ph.D student in Machine learning at Cambridge, I work on Bayesian Non-parametrics, Gaussian Processes, Kernel Learning. Application Areas: High Energy Physics, Astronomy, Science!
Alessandro Vullo (University College Dublin)
Nicolas Durrande (Secondmind)
Carl Edward Rasmussen (University of Cambridge)
More from the Same Authors
-
2021 : One Step at a Time: Pros and Cons of Multi-Step Meta-Gradient Reinforcement Learning »
Clément Bonnet · Paul Caron · Thomas D Barrett · Ian Davies · Alexandre Laterre -
2022 : Dynamic Head Pruning in Transformers »
Prisha Satwani · yiren zhao · Vidhi Lalchand · Robert Mullins -
2022 : Gaussian Process parameterized Covariance Kernels for Non-stationary Regression »
Vidhi Lalchand · Talay Cheema · Laurence Aitchison · Carl Edward Rasmussen -
2022 Poster: Sparse Gaussian Process Hyperparameters: Optimize or Integrate? »
Vidhi Lalchand · Wessel Bruinsma · David Burt · Carl Edward Rasmussen -
2021 Poster: Marginalised Gaussian Processes with Nested Sampling »
Fergus Simpson · Vidhi Lalchand · Carl Edward Rasmussen -
2021 Poster: Deep Neural Networks as Point Estimates for Deep Gaussian Processes »
Vincent Dutordoir · James Hensman · Mark van der Wilk · Carl Henrik Ek · Zoubin Ghahramani · Nicolas Durrande -
2020 : Combining variational autoencoder representations with structural descriptors improves prediction of docking scores »
Miguel Garcia-Ortegon · Carl Edward Rasmussen · Hiroshi Kajino -
2020 Poster: Ensembling geophysical models with Bayesian Neural Networks »
Ushnish Sengupta · Matt Amos · Scott Hosking · Carl Edward Rasmussen · Matthew Juniper · Paul Young -
2017 Poster: Convolutional Gaussian Processes »
Mark van der Wilk · Carl Edward Rasmussen · James Hensman -
2017 Oral: Convolutional Gaussian Processes »
Mark van der Wilk · Carl Edward Rasmussen · James Hensman -
2017 Poster: Data-Efficient Reinforcement Learning in Continuous State-Action Gaussian-POMDPs »
Rowan McAllister · Carl Edward Rasmussen -
2016 Poster: Understanding Probabilistic Sparse Gaussian Process Approximations »
Matthias Bauer · Mark van der Wilk · Carl Edward Rasmussen -
2014 Poster: Distributed Variational Inference in Sparse Gaussian Process Regression and Latent Variable Models »
Yarin Gal · Mark van der Wilk · Carl Edward Rasmussen -
2014 Poster: Variational Gaussian Process State-Space Models »
Roger Frigola · Yutian Chen · Carl Edward Rasmussen -
2013 Poster: Bayesian Inference and Learning in Gaussian Process State-Space Models with Particle MCMC »
Roger Frigola · Fredrik Lindsten · Thomas Schön · Carl Edward Rasmussen -
2012 Poster: Active Learning of Model Evidence Using Bayesian Quadrature »
Michael A Osborne · David Duvenaud · Roman Garnett · Carl Edward Rasmussen · Stephen J Roberts · Zoubin Ghahramani -
2011 Poster: Gaussian Process Training with Input Noise »
Andrew McHutchon · Carl Edward Rasmussen -
2011 Poster: Additive Gaussian Processes »
David Duvenaud · Hannes Nickisch · Carl Edward Rasmussen -
2009 Workshop: Probabilistic Approaches for Control and Robotics »
Marc Deisenroth · Hilbert J Kappen · Emo Todorov · Duy Nguyen-Tuong · Carl Edward Rasmussen · Jan Peters -
2006 Tutorial: Advances in Gaussian Processes »
Carl Edward Rasmussen