This study introduces SpectralMix, a joint dimensionality reduction technique for multi-relational graphs with categorical node attributes. SpectralMix integrates all information available from the attributes, the different types of relations, and the graph structure. It generalizes existing techniques and reduces to spectral embedding and clustering when applied to a single graph. Experiments conducted on several real-world datasets show the superiority of SpectralMix over existing methods.
Publication date: 6 Nov 2023
Project Page: https://doi.org/10.1145/3447548.3467381
Paper: https://arxiv.org/pdf/2311.01840
