PhD ceremony Mr. S. Haase: The application of divergences in prototype based vector quantization

PhD ceremony: Mr. S. Haase

Dissertation: The application of divergences in prototype based vector quantization

Promotor(s): prof. M. Biehl, prof. T. Vilmann

Faculty: Mathematics and Natural Sciences

This thesis gives a systematic analysis of machine learning relying on divergences and provides the mathematical framework for the use of these information theoretic dissimilarity measures in various learning schemes, including gradient based training prescriptions. In particular, we focus on unsupervised and supervised prototype based vector quantization as well as on dimension reduction and visualization generalizing the SNE approach.