کنفرانس بین المللی فناوری اطلاعات و دانش

English

صفحه اصلی / یازدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Improving hypergraph attention and hypergraph convolution networks

نویسندگان :

Mustafa Mohammadi Gharasuie¹ Mahmood Shabankhah² Ali Kamandi³

1- دانشگاه تهران 2- دانشگاه تهران 3- ٔدانشگاه تهران

کلمات کلیدی :

Graph Attention Network (GAT), Graph Convolutional Network (GCN), Hypergraph Attention Network (HAN), Hypergraph Convolutional Network (HGCN). a

چکیده :

Graph Neural Networks (GNNs) are models that use the structure of graphs to better exploit the bilateral relationship between neighboring nodes. Some problems, however, require that we consider a more general relationship which involve not only two nodes but rather a group of nodes. This is the approach adopted in Hypergraph Convolution and Hypergraph Attention Networks (HGAN) [1]. In this paper, we first propose to incorporate a weight matrix into these networks which, as our experimentations show, can improve the performance of the models in question. The other novelty in our work is the introduction of self-loops in the graphs which again leads to slight improvements in the accuracy of HAN. Index Terms—Graph Attention Network (GAT), Graph Convolutional Network (GCN), Hypergraph Attention Network (HAN), Hypergraph Convolutional Network (HGCN).