Timezone: »

 
Poster
High-Order Pooling for Graph Neural Networks with Tensor Decomposition
Chenqing Hua · Guillaume Rabusseau · Jian Tang

Thu Dec 01 09:00 AM -- 11:00 AM (PST) @ Hall J #217
in Poster Session 5 »

Graph Neural Networks (GNNs) are attracting growing attention due to their effectiveness and flexibility in modeling a variety of graph-structured data. Exiting GNN architectures usually adopt simple pooling operations~(\eg{} sum, average, max) when aggregating messages from a local neighborhood for updating node representation or pooling node representations from the entire graph to compute the graph representation. Though simple and effective, these linear operations do not model high-order non-linear interactions among nodes. We propose the Tensorized Graph Neural Network (tGNN), a highly expressive GNN architecture relying on tensor decomposition to model high-order non-linear node interactions. tGNN leverages the symmetric CP decomposition to efficiently parameterize permutation-invariant multilinear maps for modeling node interactions. Theoretical and empirical analysis on both node and graph classification tasks show the superiority of tGNN over competitive baselines. In particular, tGNN achieves the most solid results on two OGB node classification datasets and one OGB graph classification dataset.

Keywords: graph neural networks · Graph classification · Tensor · Node Classification · CP decomposition
[ Paper]  [ Poster]  [ OpenReview] 

Author Information

Chenqing Hua (McGill University)
Guillaume Rabusseau (Mila - Université de Montréal)
Jian Tang (Mila)

More from the Same Authors