Finite rank representations of a Gaussian Process
In this video we will discuss an alternative representation of Gaussian Processes using the the Karhunen-Loève expansion (KL). We will provide a visual explanation of this expansion and discuss how we can obtain a finite-rank approximation of the Gaussian Process.
Thanks to this approximation we can obtain algorithms that can deal with shapes that are represented using millions of points. Furthermore, the representation gives us a simple and efficient means for computing the probability of any shape in the family.
3-6: Finite rank representation of a Gaussian process from NMC Universität Basel on Vimeo.