Uplifting Message Passing Neural Network with Graph Original Information
This work addresses the problem of improving graph neural network performance for node classification tasks, offering a novel method that is incremental in enhancing existing MPNN frameworks.
The paper tackled the limited expressive power and accuracy of message passing neural networks (MPNNs) in node classification by proposing INGNN, a model that fully exploits graph original information, achieving an average rank of 1.78 compared to state-of-the-art methods.
Message passing neural networks (MPNNs) learn the representation of graph-structured data based on graph original information, including node features and graph structures, and have shown astonishing improvement in node classification tasks. However, the expressive power of MPNNs is upper bounded by the first-order Weisfeiler-Leman test and its accuracy still has room for improvement. This work studies how to improve MPNNs' expressiveness and generalizability by fully exploiting graph original information both theoretically and empirically. It further proposes a new GNN model called INGNN (INformation-enhanced Graph Neural Network) that leverages the insights to improve node classification performance. Extensive experiments on both synthetic and real datasets demonstrate the superiority (average rank 1.78) of our INGNN compared with state-of-the-art methods.