#eth/hpc

Hypergraph Neural Networks Paper

Hypergraph neural networks framework
encodes high-order data correlation in a hypergraph structure
hypergraphs are employed in many computer vision tasks such as classification and retrieval tasks
Learning methods suffer from high computation complexity and storage costs

HGNN Proposal

A hypergraph is defined as:

$G = (V, E, W)$
vertex set $V$
hyperedge set $E$
weight matrix $W$ , each hyperedge is assigned with a weight
incidence matrix $H$ defines $G$ , dimensions $| V | \times | E |$
$h (v, e) = {\begin{cases} 1, & if v \in e \\ 0, & if v \notin e \end{cases}$
weight function for vertex $v$ is $d (v) = \sum_{e \in E} w (e) h (v, e)$
weight function for edge $e$ is $d (e) = \sum_{v \in V} h (v, e)$
$D_{e}$ and $D_{v}$ diagonal matrices of edge and vertex degrees