[WWW`20]MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding

[Peiyance]

The main goal of reading paper is not just understanding it. Try to understand the key concept, but we need to get new ideas and research directions from the paper.

Paper information

• title: MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding
• authors: Xinyu Fu, Jiani Zhang, Ziqiao Meng, Irwin King
• venue: WWW 2020
• pdf link: link
• github: link
• abstract: A large number of real-world graphs or networks are inherently het- erogeneous, involving a diversity of node types and relation types. Heterogeneous graph embedding is to embed rich structural and se- mantic information of a heterogeneous graph into low-dimensional node representations. Existing models usually define multiple meta- paths in a heterogeneous graph to capture the composite relations and guide neighbor selection. However, these models either omit node content features, discard intermediate nodes along the meta- path, or only consider one metapath. To address these three lim- itations, we propose a new model named Metapath Aggregated Graph Neural Network (MAGNN) to boost the final performance. Specifically, MAGNN employs three major components, i.e., the node content transformation to encapsulate input node attributes, the intra-metapath aggregation to incorporate intermediate seman- tic nodes, and the inter-metapath aggregation to combine mes- sages from multiple metapaths. Extensive experiments on three real-world heterogeneous graph datasets for node classification, node clustering, and link prediction show that MAGNN achieves more accurate prediction results than state-of-the-art baselines.

Summary: problems to address, key ideas, quick results

presentation link

Questions about the paper?

What do you like?

The overall idea of this paper is relatively clear, focusing on the metapath of heterogeneous graph, and aiming at the two kinds of relationships of intra- or inter-metapath, an aggregation idea based on attention is proposed.

What you don't like?

How to improve?

Any new ideas?

As a heterogeneous graph is defined as a graph which has various node types and edge types. This paper only considers the situation where the graph has multiple node types but only one edge type. Maybe how to handle the problem where different edge types existed will be the next direction.

Reproducing results (if any)