Graph Automorphism Group Equivariant Neural Networks

Authors: Edward Pearce-Crump, William Knottenbelt

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical This paper presents work that is primarily a theoretical contribution; hence we do not expect profound societal impact in the short term. However, in the medium term, a number of applications may well emerge from the theory having high levels of impact. We are the first to show how the combinatorics underlying bilabelled graphs provides the theoretical background for constructing neural networks that are equivariant to the automorphism group of a graph having n vertices where the layers are some tensor power of Rn.
Researcher Affiliation Academia Edward Pearce Crump 1 William J. Knottenbelt 1 1Department of Computing, Imperial College London, United Kingdom. Correspondence to: Edward Pearce Crump <ep1011@ic.ac.uk>.
Pseudocode Yes Procedure: Weight Matrix for an Aut(G)-Equivariant Linear Layer Function from (Rn) k to (Rn) l.
Open Source Code No The paper does not provide any explicit statements or links indicating the release of open-source code for the described methodology.
Open Datasets No The paper is theoretical and does not involve experimental evaluation on datasets, thus no dataset availability information is provided.
Dataset Splits No The paper is theoretical and does not involve experimental evaluation, thus no training/validation/test splits are specified.
Hardware Specification No The paper is theoretical and does not report specific experimental hardware used for its analysis or derivations.
Software Dependencies No The paper is theoretical and does not describe software dependencies with specific version numbers.
Experiment Setup No The paper is theoretical and describes a mathematical procedure rather than an experimental setup with hyperparameters or system-level training settings.