Generalization and Representational Limits of Graph Neural Networks
arXiv:2002.06157
Abstract
We address two fundamental questions about graph neural networks (GNNs). First, we prove that several important graph properties cannot be computed by GNNs that rely entirely on local information. Such GNNs include the standard message passing models, and more powerful spatial variants that exploit local graph structure (e.g., via relative orientation of messages, or local port ordering) to distinguish neighbors of each node. Our treatment includes a novel graph-theoretic formalism. Second, we provide the first data dependent generalization bounds for message passing GNNs. This analysis explicitly accounts for the local permutation invariance of GNNs. Our bounds are much tighter than existing VC-dimension based guarantees for GNNs, and are comparable to Rademacher bounds for recurrent neural networks.
References in corpus (6)
- Interaction Networks for Learning about Objects, Relations and Physics
- Graph Transformer Networks
- Learning Multimodal Graph-to-Graph Translation for Molecular Optimization
- Position-aware Graph Neural Networks
- Efficient Probabilistic Logic Reasoning with Graph Neural Networks
- On Generalization Bounds of a Family of Recurrent Neural Networks