Interpolating Graph Pair to Regularize Graph Classification
Authors: Hongyu Guo, Yongyi Mao
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments, using eight benchmarking tasks from various domains, showing that our strategy can effectively regularize the graph classification to improve its predictive accuracy, outperforming popular graph augmen- tation approaches and GNN methods. |
| Researcher Affiliation | Collaboration | Hongyu Guo1,2, Yongyi Mao2 1National Research Council Canada 2University of Ottawa {hongyu.guo, ymao}@uottawa.ca |
| Pseudocode | Yes | Algorithm 1: The mixing schema in if Mixup |
| Open Source Code | No | The paper does not provide a link or explicit statement about the availability of its own source code. |
| Open Datasets | Yes | We use eight graph classification tasks from the graph benchmark datasets collection TUDatasets (Morris et al. 2020)... and can be downloaded directly using Py Torch Geometric (Fey and Lenssen 2019) s build-in function online 1. 1https://chrsmrrs.github.io/datasets/docs/datasets |
| Dataset Splits | Yes | We evaluate the models using 10-fold cross validation, and report the mean and standard deviation of three runs on a cluster with GPU nodes of NVidia V100 with 32 GB memory. |
| Hardware Specification | Yes | We report mean and standard deviation of three runs on a cluster with GPU nodes of NVidia V100 with 32 GB memory. |
| Software Dependencies | No | The paper mentions software like PyTorch Geometric, Adam W optimizer, GCN, and GIN, but does not provide specific version numbers for these dependencies, which is required for reproducibility. |
| Experiment Setup | Yes | The hyper-parameters we search for all models on each dataset are as follows: (1) initial learning rate {0.01, 0.0005}; (2) hidden unit of size {64, 128}; (3) batch size {32, 128}; (4) dropout ratio after the dense layer {0, 0.5}; (5) Drop Node and Drop Edge drop ratio {20%, 40%}; (6) number of layers in GNNs {3, 5, 8}; (7) Beta distribution for if Mixup, Mixup Graph and Manifold Mixup {Beta(1, 1), Beta(2, 2), Beta(20, 1)}. |