Generalization and Representational Limits of Graph Neural Networks

Authors: Vikas Garg, Stefanie Jegelka, Tommi Jaakkola

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical First, we prove that several important graph properties, e.g., shortest/longest cycle, diameter, or certain motifs, cannot be computed by GNNs that rely entirely on local information. 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. We outline some key proof steps for our results in the main text, and defer the detailed proofs to the Supplementary material.
Researcher Affiliation Academia 1CSAIL, MIT. Correspondence to: Vikas Garg <vgarg@csail.mit.edu>, Stefanie Jegelka <stefje@mit.edu>, Tommi Jaakkola <tommi@csail.mit.edu>.
Pseudocode No The paper provides theoretical analysis, propositions, and lemmas, but does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements about releasing source code or provide links to a code repository for the methodology described.
Open Datasets No The paper discusses theoretical limits and generalization bounds of GNNs and does not describe using or providing access information for a specific public dataset for its own research or experiments.
Dataset Splits No The paper focuses on theoretical analysis and does not include details on dataset splits for training, validation, or testing.
Hardware Specification No The paper does not describe any experimental setup that would require specific hardware, and thus no hardware specifications are mentioned.
Software Dependencies No The paper is theoretical and does not mention any specific software dependencies with version numbers required for reproducibility.
Experiment Setup No The paper is theoretical and does not detail any experimental setup, hyperparameters, or training configurations.