Autobahn: Automorphism-based Graph Neural Nets
Authors: Erik Thiede, Wenda Zhou, Risi Kondor
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present empirical results from an implementation of our architecture on two subsets of the ZINC dataset using the data splits from [23], as well as three standardized tasks from Open Graph Benchmark. All datasets are released under the MIT license. Baselines were taken from [21, 23, 47, 48]. Our automorphism-based neural network achieves results competitive with modern MPNNs. and Table 1: Performance of our Autobahn architecture on two splits of the ZINC dataset and three datasets in the OGB benchmark family, compared with other recent message passing architectures. ZINC experiments use MAE (lower is better); for all other metrics higher is better. |
| Researcher Affiliation | Academia | 1 Center for Computational Mathematics, Flatiron Institute, New York NY 10010 2 Center for Data Science, New York University, New York NY 10011 3 Department of Computer Science, University of Chicago, Chicago IL 60637 4 Department of Statistics, University of Chicago, Chicago IL 60637 |
| Pseudocode | Yes | Algorithm 1 Automorphism-based Neuron |
| Open Source Code | Yes | Full details of the model, including training hyper-parameters and architecture details, are available in the Supplement; code is freely available at https://github.com/risilab/Autobahn. |
| Open Datasets | Yes | We present empirical results from an implementation of our architecture on two subsets of the ZINC dataset using the data splits from [23], as well as three standardized tasks from Open Graph Benchmark. All datasets are released under the MIT license. |
| Dataset Splits | Yes | We present empirical results from an implementation of our architecture on two subsets of the ZINC dataset using the data splits from [23], as well as three standardized tasks from Open Graph Benchmark. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications) used for running experiments. |
| Software Dependencies | No | The paper does not provide specific software dependency versions (e.g., library names with version numbers like PyTorch 1.9). |
| Experiment Setup | No | Full details of the model, including training hyper-parameters and architecture details, are available in the Supplement (implying they are not in the main paper body). |