Learning From Complex Networks
MCML Authors
Christian Frey
Dr.
* Former Member
Abstract
Christian Frey
Dr.
* Former Member
Abstract
This thesis addresses key challenges in modern graph-based applications by proposing advanced techniques in spectral clustering, graph neural networks, and probabilistic graph structures. It introduces a robust, accelerated spectral clustering model for homogeneous graphs and a transformer-inspired Graph Shell Attention model to counter over-smoothing in graph neural networks. Furthermore, it tackles optimization in uncertain networks, presents a new approach to a vehicle routing problem with flexible delivery locations, and provides a novel method for classifying social media trends, illustrating the vital role of AI in understanding complex graph structures. (Shortened).
BibTeXKey: Fre23