Deep kernel methods
Thursday, November 25, 2021
13:00 UTC
Dr Laurence Aitchison
University of Bristol
Recording
Open recording
Hosted by Sheffield ML70 minutes
Abstract
Deep neural networks (DNNs) with the flexibility to learn good top-layer representations have eclipsed shallow kernel methods without that flexibility. Here, we take inspiration from deep neural networks to develop a new family of deep kernel method. In a deep kernel method, there is a kernel at every layer, and the kernels are jointly optimized to improve performance (with strong regularisation). We establish the representational power of deep kernel methods, by showing that they perform exact inference in an infinitely wide Bayesian neural network or deep Gaussian process. Next, we conjecture that the deep kernel machine objective is unimodal, and give a proof of unimodality for linear kernels. Finally, we exploit the simplicity of the deep kernel machine loss to develop a new family of optimizers, based on a matrix equation from control theory, that converges in around 10 steps.
Topics
Related Job Opportunities
JobJobJob
Postdoctoral Scientist - Sarvestani Lab
The Sarvestani Lab at Cornell University is recruiting a postdoctoral scientist in systems neuroscience to study how visual and motor systems across the brain and body support perception and…
Posted Aug 12, 2026
Ithaca, New York, United States
PhD Studentship: Mitochondrial Metabolism and Novel Therapeutic Strategies for Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) (Fixed Term)
Supervisors: Professor Andrew Murray, Department of Physiology, Development and Neuroscience, University of Cambridge Dr Ross Lindsay, Novo Nordisk Funding: Fully funded PhD studentship (Home/UK…
Posted Aug 10, 2026
Research Associate (Fixed Term)
We seek a highly motivated Postdoctoral Research Associate to join the laboratory of Professor Kathy Niakan. We are based in the Loke Centre for Trophoblast Research (LCTR), in the Department of…
Posted Aug 11, 2026