The goal of this work is to analyze the differences in performance between the new generation and the universal Graph Neural Network using traditional Fourier decomposition and to propose an upper bound on the approximation error. In addition, the Graph Neural Network's ability to characterize hierarchical data will be investigated and practically evaluated on a knowledge graph benchmarks. The contents of the master’s thesis are already well defined. Initial code and data will be provided to the student.
Further information can be found in the attached PDF.
Ausschreibung: Download (pdf)