Graph Mixup with Soft Alignments

Authors: Hongyi Ling, Zhimeng Jiang, Meng Liu, Shuiwang Ji, Na Zou

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct systematic experiments to show that S-Mixup can improve the performance and generalization of graph neural networks (GNNs) on various graph classification tasks.
Researcher Affiliation Academia 1Department of Computer Science & Engineering, Texas A&M University, TX, USA 2Department of Engineering Technology & Industrial Distribution, Texas A&M University, TX, USA.
Pseudocode Yes Algorithm 1 Training algorithm", "Algorithm 2 Mixup algorithm
Open Source Code Yes Our code is publicly available as part of the DIG package (https://github.com/divelab/DIG).
Open Datasets Yes In this section, we evaluate the effectiveness of our method on six real-world datasets from the TUDatasets benchmark (Morris et al., 2020)2... We also conduct experiments on ogbg-molhiv, which is a large molecular graph dataset from OGB benchmark (Hu et al., 2020)3.
Dataset Splits Yes For the TUDatasetes benchmark, we randomly split the dataset into train/validation/test data by 80%/10%/10%.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper mentions using GCN and GIN models and the Adam optimizer, but does not provide specific software dependencies with version numbers (e.g., Python, PyTorch/TensorFlow, specific graph neural network libraries).
Experiment Setup Yes We use the Adam optimizer (Kingma & Ba, 2015) to train all models. See Table 8 for the hyperparameters of training the classification model. ... For the graph matching network used in S-Mixup, we set the hidden size as 256 and the readout layer as global sum pooling. For all six datasets, the graph matching network is trained for 500 epochs with a learning rate of 0.001. For the number of layers and batch size, see Table 9.