Iterative Hypergraph Computation Based on Hyperedge-Connected Graphs
Abstract: A hypergraph allows a hyperedge to connect arbitrary number of vertices, which can be used to capture the complex and high-order relationships. By analyzing the iterative processing on bipartite graphs, a method of converting the original hypergraph into a hyperedge-connected graph and corresponding iterative processing method are proposed. Then, the iterative processing solution based on hyperedge-connected graphs is combined with Push-based and Pull-based message acquisition mechanisms. On top of the distributed graph processing system HybridGraph, a hypergraph iterative processing framework HyraphD is implemented. Finally, extensive experiments are conducted on several real-world datasets and hypergraph learning algorithms. Experimental results confirm the efficiency and the scalability of HyraphD.
Loading