Decoupling the Depth and Scope of Graph Neural Networks
Authors: Hanqing Zeng, Muhan Zhang, Yinglong Xia, Ajitesh Srivastava, Andrey Malevich, Rajgopal Kannan, Viktor Prasanna, Long Jin, Ren Chen
NeurIPS 2021 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, on seven graphs (with up to 110M nodes) and six backbone GNN architectures, our design achieves significant accuracy improvement with orders of magnitude reduction in computation and hardware cost. |
| Researcher Affiliation | Collaboration | Hanqing Zeng USC EMAIL Muhan Zhang Peking University, BIGAI EMAIL Yinglong Xia Facebook AI EMAIL Ajitesh Srivastava USC EMAIL Andrey Malevich Facebook AI EMAIL Rajgopal Kannan US ARL EMAIL Viktor Prasanna USC EMAIL Long Jin Facebook AI EMAIL Ren Chen Facebook AI EMAIL |
| Pseudocode | Yes | See Appendix D and F.3 for algorithm and experiments. |
| Open Source Code | Yes | Our code is available at https://github.com/facebookresearch/shaDow_GNN |
| Open Datasets | Yes | We evaluate SHADOW-GNN on seven graphs. Six of them are for the node classification task: Flickr [55], Reddit [12], Yelp [55], ogbn-arxiv, ogbn-products and ogbn-papers100M [16]. |
| Dataset Splits | Yes | We follow the default data splits for all datasets, which are usually 60% training, 20% validation, and 20% test. For ogbn-papers100M, the training, validation, test splits are 80%, 10%, 10% respectively. (from Appendix E.1) |
| Hardware Specification | No | Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [No] |
| Software Dependencies | No | The paper mentions using W&B [5] but does not provide specific version numbers for software dependencies or libraries in the text. |
| Experiment Setup | Yes | All models on all datasets have uniform hidden dimension of 256. [...] For the model depth, since L = 3 is the standard setting in the literature (e.g., see the benchmarking in OGB [16]), we start from L = 3 and further evaluate a deeper model of L = 5. Hyperparameter tuning and architecture configurations are in Appendix E.4. (Appendix E.4 specifies: 'hidden dimension of 256 for all models', 'learning rate of 0.001', 'Adam optimizer [21] with weight decay of 5e-4', 'number of training epochs is 1000', 'dropout rate set to 0.5') |