Are Powerful Graph Neural Nets Necessary? A Dissection on Graph Classification

Abstract

Graph Neural Nets (GNNs) have received increasing attentions, partially due to their superior performance in many node and graph classification tasks. However, there is a lack of understanding on what they are learning and how sophisticated the learned graph functions are. In this work, we propose a dissection of GNNs on graph classification into two parts: 1) the graph filtering, where graph-based neighbor aggregations are performed, and 2) the set function, where a set of hidden node features are composed for prediction. To study the importance of both parts, we propose to linearize them separately. We first linearize the graph filtering function, resulting Graph Feature Network (GFN), which is a simple lightweight neural net defined on a \textit{set} of graph augmented features. Further linearization of GFN's set function results in Graph Linear Network (GLN), which is a linear function. Empirically we perform evaluations on common graph classification benchmarks. To our surprise, we find that, despite the simplification, GFN could match or exceed the best accuracies produced by recently proposed GNNs (with a fraction of computation cost), while GLN underperforms significantly. Our results demonstrate the importance of non-linear set function, and suggest that linear graph filtering with non-linear set function is an efficient and powerful scheme for modeling existing graph classification benchmarks.

arXiv commentA shorter version titled "Graph Feature Networks" was accepted to ICLR'19 RLGM workshop. code available at https://github.com/chentingpc/gfn

References in corpus (14)

• Adam: A Method for Stochastic Optimization
• Semi-Supervised Classification with Graph Convolutional Networks
• Inductive Representation Learning on Large Graphs
• Neural Message Passing for Quantum Chemistry
• Fast Graph Representation Learning with PyTorch Geometric
• Simplifying Graph Convolutional Networks
• Representation Learning on Graphs with Jumping Knowledge Networks
• Learning Convolutional Neural Networks for Graphs
• Discriminative Embeddings of Latent Variable Models for Structured Data
• Gated Graph Sequence Neural Networks
• Revisiting Graph Neural Networks: All We Have is Low-Pass Filters
• LanczosNet: Multi-Scale Deep Graph Convolutional Networks
• Anonymous Walk Embeddings
• Understanding the Representation Power of Graph Neural Networks in Learning Graph Topology

Cited by in corpus (14)

• Graph Contrastive Learning with Augmentations
• A Comprehensive Survey on Deep Graph Representation Learning
• TUDataset: A collection of benchmark datasets for learning with graphs
• SimGRACE: A Simple Framework for Graph Contrastive Learning without Data Augmentation
• Benchmarking Graph Neural Networks
• Rethinking pooling in graph neural networks
• Scalable Graph Neural Networks for Heterogeneous Graphs
• Graph Neural Networks with Local Graph Parameters
• Evaluating Logical Generalization in Graph Neural Networks
• On Graph Neural Networks versus Graph-Augmented MLPs
• Quaternion Graph Neural Networks
• Ego-based Entropy Measures for Structural Representations
• A Structural Feature-Based Approach for Comprehensive Graph Classification
• Improving Expressivity of Graph Neural Networks