Semi-supervised Hypergraph Node Classification on Hypergraph Line Expansion
Semi-supervised Hypergraph Node Classification on Hypergraph Line Expansion
复制标题
DOI：
10.1145/3511808.3557447
复制
发表时间：
2020-05
期刊：
影响因子：
--
通讯作者：
Chaoqi Yang;Ruijie Wang;Shuochao Yao;T. Abdelzaher
中科院分区：
文献类型：
--
作者：
Chaoqi Yang;Ruijie Wang;Shuochao Yao;T. Abdelzaher
以前的超图扩展仅在顶点级别或超边级别上进行，从而忽略了数据共现的对称性，并导致信息丢失。为了解决这个问题，本文平等对待顶点和超边，并提出了一种新的超图扩展，称为线扩展（LE），用于超图学习。新的展开通过对顶点-超边对进行建模，双射地从超图导出同质结构。我们的建议本质上将超图简化为简单图，这使得现有的图学习算法能够与高阶结构无缝协作。我们进一步证明我们的线扩展是各种超图扩展的统一框架。我们根据超图节点分类任务在五个超图数据集上评估所提出的 LE。结果表明，我们的方法可以始终比最佳基线提高至少 2% 的准确度。
Previous hypergraph expansions are solely carried out on either vertex level or hyperedge level, thereby missing the symmetric nature of data co-occurrence, and resulting in information loss. To address the problem, this paper treats vertices and hyperedges equally and proposes a new hypergraph expansion named the line expansion(LE) for hypergraphs learning. The new expansion bijectively induces a homogeneous structure from the hypergraph by modeling vertex-hyperedge pairs. Our proposal essentially reduces the hypergraph to a simple graph, which enables the existing graph learning algorithms to work seamlessly with the higher-order structure. We further prove that our line expansion is a unifying framework over various hypergraph expansions. We evaluate the proposed LE on five hypergraph datasets in terms of the hypergraph node classification task. The results show that our method could achieve at least 2% accuracy improvement over the best baseline consistently.