We introduce scalable deep kernels, which combine the structural properties
of deep learning architectures with the non-parametric flexibility of kernel
methods. Specifically, we transform the inputs of a spectral mixture base
kernel with a deep architecture, using local kernel interpolation, inducing
points, and structure exploiting (Kronecker and Toeplitz) algebra for a
scalable kernel representation. These closed-form kernels can be used as
drop-in replacements for standard kernels, with benefits in expressive power
and scalability. We jointly learn the properties of these kernels through the
marginal likelihood of a Gaussian process. Inference and learning cost $O(n)$
for $n$ training points, and predictions cost $O(1)$ per test point. On a large
and diverse collection of applications, including a dataset with 2 million
examples, we show improved performance over scalable Gaussian processes with
flexible kernel learning models, and stand-alone deep architectures.
%0 Generic
%1 wilson2015kernel
%A Wilson, Andrew Gordon
%A Hu, Zhiting
%A Salakhutdinov, Ruslan
%A Xing, Eric P.
%D 2015
%K deep-learning from:adulny gaussian-processes kernel-learning non-parametric
%T Deep Kernel Learning
%U http://arxiv.org/abs/1511.02222
%X We introduce scalable deep kernels, which combine the structural properties
of deep learning architectures with the non-parametric flexibility of kernel
methods. Specifically, we transform the inputs of a spectral mixture base
kernel with a deep architecture, using local kernel interpolation, inducing
points, and structure exploiting (Kronecker and Toeplitz) algebra for a
scalable kernel representation. These closed-form kernels can be used as
drop-in replacements for standard kernels, with benefits in expressive power
and scalability. We jointly learn the properties of these kernels through the
marginal likelihood of a Gaussian process. Inference and learning cost $O(n)$
for $n$ training points, and predictions cost $O(1)$ per test point. On a large
and diverse collection of applications, including a dataset with 2 million
examples, we show improved performance over scalable Gaussian processes with
flexible kernel learning models, and stand-alone deep architectures.
@misc{wilson2015kernel,
abstract = {We introduce scalable deep kernels, which combine the structural properties
of deep learning architectures with the non-parametric flexibility of kernel
methods. Specifically, we transform the inputs of a spectral mixture base
kernel with a deep architecture, using local kernel interpolation, inducing
points, and structure exploiting (Kronecker and Toeplitz) algebra for a
scalable kernel representation. These closed-form kernels can be used as
drop-in replacements for standard kernels, with benefits in expressive power
and scalability. We jointly learn the properties of these kernels through the
marginal likelihood of a Gaussian process. Inference and learning cost $O(n)$
for $n$ training points, and predictions cost $O(1)$ per test point. On a large
and diverse collection of applications, including a dataset with 2 million
examples, we show improved performance over scalable Gaussian processes with
flexible kernel learning models, and stand-alone deep architectures.},
added-at = {2021-12-13T10:02:12.000+0100},
author = {Wilson, Andrew Gordon and Hu, Zhiting and Salakhutdinov, Ruslan and Xing, Eric P.},
biburl = {https://www.bibsonomy.org/bibtex/26ed29ff479b53e1d91e2636d4b2acdd4/adulny},
description = {Deep Kernel Learning},
interhash = {17afb930bd2af0bcda42212d229ff7d9},
intrahash = {6ed29ff479b53e1d91e2636d4b2acdd4},
keywords = {deep-learning from:adulny gaussian-processes kernel-learning non-parametric},
note = {cite arxiv:1511.02222Comment: 19 pages, 6 figures},
timestamp = {2021-12-13T10:02:12.000+0100},
title = {Deep Kernel Learning},
url = {http://arxiv.org/abs/1511.02222},
year = 2015
}