Goal: minimize unknown function $\phi$ in as few evaluations as possible.
Automatic explore-exploit tradeoff.
$$ \htmlData{fragment-index=0,class=fragment}{ x_0 } \qquad \htmlData{fragment-index=1,class=fragment}{ x_1 = x_0 + f(x_0)\Delta t } \qquad \htmlData{fragment-index=2,class=fragment}{ x_2 = x_1 + f(x_1)\Delta t } \qquad \htmlData{fragment-index=3,class=fragment}{ .. } $$
$\nu = 1/2$
$\nu = 3/2$
$\nu = 5/2$
$\nu = \infty$
$$
\htmlData{class=fragment fade-out,fragment-index=9}{
\footnotesize
\mathclap{
k_\nu(x,x') = \sigma^2 \frac{2^{1-\nu}}{\Gamma(\nu)} \del{\sqrt{2\nu} \frac{\norm{x-x'}}{\kappa}}^\nu K_\nu \del{\sqrt{2\nu} \frac{\norm{x-x'}}{\kappa}}
}
}
\htmlData{class=fragment d-print-none,fragment-index=9}{
\footnotesize
\mathclap{
k_\infty(x,x') = \sigma^2 \exp\del{-\frac{\norm{x-x'}^2}{2\kappa^2}}
}
}
$$
$\sigma^2$: variance
$\kappa$: length scale
$\nu$: smoothness
$\nu\to\infty$: recovers squared exponential kernel
$$ k_\infty^{(d_g)}(x,x') = \sigma^2\exp\del{-\frac{d_g(x,x')^2}{2\kappa^2}} $$
Theorem. (Feragen et al.) Let $M$ be a complete Riemannian manifold without boundary. If $k_\infty^{(d_g)}$ is positive semi-definite for all $\kappa$, then $M$ is isometric to a Euclidean space.
$$ \htmlData{class=fragment,fragment-index=0}{ \underset{\t{Matérn}}{\undergroup{\del{\frac{2\nu}{\kappa^2} - \Delta}^{\frac{\nu}{2}+\frac{d}{4}} f = \c{W}}} } \qquad \htmlData{class=fragment,fragment-index=1}{ \underset{\t{squared exponential}}{\undergroup{\vphantom{\del{\frac{2\nu}{\kappa^2} - \Delta}^{\frac{\nu}{2}+\frac{d}{4}}} e^{-\frac{\kappa^2}{4}\Delta} f = \c{W}}} } $$ $\Delta$: Laplacian $\c{W}$: (rescaled) white noise
$$ k_\nu(x,x') = \frac{\sigma^2}{C_\nu} \sum_{n=0}^\infty \del{\frac{2\nu}{\kappa^2} - \lambda_n}^{\nu-\frac{d}{2}} f_n(x) f_n(x') $$
V. Borovitskiy,* P. Mostowsky,* A. Terenin,* M. P. Deisenroth. Matérn Gaussian Processes on Riemannian Manifolds. Advances in Neural Information Processing Systems, 2020.
N. Jaquier,* V. Borovitskiy,* A. Smolensky, A. Terenin, T. Asfour, L. Rozo. Geometry-aware Bayesian Optimization in Robotics using Riemannian Matérn Kernels. Conference on Robot Learning, 2021.
V. Borovitskiy,* I. Azangulov,* P. Mostowsky,* A. Terenin,* M. P. Deisenroth, N. Durrande. Matérn Gaussian Processes on Graphs. Artificial Intelligence and Statistics, 2021.
M. J. Hutchinson,* A. Terenin,* V. Borovitskiy,* S. Takao,* Y. W. Teh, M. P. Deisenroth. Vector-valued Gaussian Processes on Riemannian Manifolds via Gauge Independent Projected Kernels. Advances in Neural Information Processing Systems, 2021.
J. T. Wilson,* V. Borovitskiy,* P. Mostowsky,* A. Terenin,* M. P. Deisenroth. Efficiently Sampling Functions from Gaussian Process Posteriors. International Conference on Machine Learning, 2020.
J. T. Wilson,* V. Borovitskiy,* P. Mostowsky,* A. Terenin,* M. P. Deisenroth. Pathwise Conditioning of Gaussian Process. Journal of Machine Learning Research, 2021.
*Equal contribution