With this ML Session series we intend to provide individual and independent lecture sessions on ML related topics.This time Prof. Dr. Marc Toussaint will talk about "Gaussian Processes".
The lecture will introduce to Gaussian Processes (GP) and GP classification from two different perspectives: As the kernelization of Bayesian ridge regression (also call "weight-space view"), and from directly formulating a prior over predictive functions (also called "function-space view"). Same for logistic regression and GP classification. The default use case of GPs is to quantify uncertainty about the predictor. I will discuss the two additional important use cases: active learning and Bayesian Optimization. Depending on time, I will briefly discuss recent deep GP formulations.