TU Berlin

Methoden der Künstlichen IntelligenzMasterarbeiten

Inhalt des Dokuments

zur Navigation

Masterarbeitensvorschlag

Logistic-Heaviside Multi-Class Gaussian Process Classification

You will have to develop a Gaussian process model based on the paper  "Scalable Multi-Class Gaussian Process Classification using Expectation Propagation". The idea is to change the approximation in order to make the inference much more scalable, by using the logistic link and the Polya-Gamma augmentation. You should be familiar with Gaussian Processes and Variational Inference. For more information contact Theo Galy-Fajou

Navigation

Direktzugang

Schnellnavigation zur Seite über Nummerneingabe