An Analysis of How Hypergraph Spectral Clustering Deals with Higher-order Relationships

Spectral clustering has recently been extended to hypergraphs, which are formed by vertices and higher-order relationships between these vertices. In this article, we consider a methodology to qualitatively compare hypergraph spectral clustering against the classical graph spectral clustering. More precisely, we use a graph representation to create a graph from a given hypergraph, thus allowing a comparison between spectral clustering methods. Experiment shows that the hypergraph spectral clustering deals differently with higher-order relationships.