We propose neighborhood-based core decomposition: a novel way of decomposing hypergraphs into hierarchical neighborhood-cohesive subhypergraphs. Alternative approaches to decomposing hypergraphs, e.g., reduction to clique or bipartite graphs, are not meaningful in certain applications, the later also results in inefficient decomposition; while existing degree-based hypergraph decomposition does not distinguish nodes with different neighborhood sizes. Our case studies show that the proposed decomposition is more effective than degree and clique graph-based decompositions in disease intervention and in extracting provably approximate and application-wise meaningful densest subhypergraphs. We propose three algorithms: Peel, its efficient variant E-Peel, and a novel local algorithm: Local-core with parallel implementation. Our most efficient parallel algorithm Local-core(P) decomposes hypergraph with 27M nodes and 17M hyperedges in-memory within 91 seconds by adopting various optimizations. Finally, we develop a new hypergraph-core model, the (neighborhood, degree)-core by considering both neighborhood and degree constraints, design its decomposition algorithm Local-core+Peel, and demonstrate its superiority in spreading diffusion.
翻译:我们提出了基于邻域的核心分解:一种将超图分解为层次化邻域内聚的子超图的新方法。替代超图分解的方法,例如转化为团或二分图,在某些应用中并没有意义,后者也导致了低效的分解;而现有的基于度数的超图分解并不能区分邻域大小不同的节点。我们的案例研究表明,所提出的分解方法在疾病干预和提取可证明近似和应用上有意义的密集子超图方面比度数和团图基础分解更有效。我们提出了三个算法:Peel算法,其中包括其有效变体E-Peel算法,以及一种新型的本地算法:Local-core并支持并行实现。我们最有效的并行算法Local-core(P)通过采用各种优化,在内存中分解包含27M个节点和17M个超边的超图,仅需91秒。最后,我们基于邻域和度约束设计了新的超图核心模型(neighborhood, degree)-core,设计了其分解算法(Local-core+Peel),并证明了其在扩散传播方面的优越性。