Nearest Neighbor Gaussian Processes
In a k-Nearest Neighbor Gaussian Process, we assume that the input points $x$ are ordered in such a way that $f(x_i)$ is independent of $f(x_j)$ whenever…
In a k-Nearest Neighbor Gaussian Process, we assume that the input points $x$ are ordered in such a way that $f(x_i)$ is independent of $f(x_j)$ whenever…
The $i$th Krylov subspace $\mathcal{K}_i$ for a symmetric matrix $A$ is the subspace spanned by repeatedly multiplying $A$ by an initial vector $b$. This mos...
The Bayesian perspective offers a simple and elegant framework for tracking uncertainty in online optimization problems. In structured bandit settings, it na...
A mean field analysis of the gossip protocol shows that distributed consensus with CRDTs can happen in constant expected time.
Instead of representing the joint distribution of an object and landmarks’ locations with a multivariate Gaussian, we can use a particle filter.
A hands-on introduction to an old technique for simultaneous localization and mapping.
This post is about a technique that allows us to use variational message passing on models where the likelihood doesn’t have a conjugate prior. There will be...
Generative image models based on ordinary differential equations can be seen as forms of variational auto-encoders.