Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Graph Mixup on Approximate Gromov–Wasserstein Geodesics
Authors: Zhichen Zeng, Ruizhong Qiu, Zhe Xu, Zhining Liu, Yuchen Yan, Tianxin Wei, Lei Ying, Jingrui He, Hanghang Tong
ICML 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that the proposed GEOMIX promotes the generalization and robustness of GNN models. |
| Researcher Affiliation | Academia | 1University of Illinois Urbana-Champaign 2University of Michigan, Ann Arbor. |
| Pseudocode | Yes | Algorithm 1 GEOMIX |
| Open Source Code | Yes | Code and datasets are available at https://github.com/zhichenz98/Geo Mix-ICML24. |
| Open Datasets | Yes | All the real-world datasets used in the paper are from (Morris et al., 2020) and available online1. We give a brief introduction to the datasets as follows PROTEINS (Borgwardt et al., 2005), MUTAG (Kriege & Mutzel, 2012), MSRC-9 (Neumann et al., 2016), IMDB-B (Yanardag & Vishwanathan, 2015), and IMDB-M (Yanardag & Vishwanathan, 2015). |
| Dataset Splits | Yes | We split the dataset into train/test/validation set by 80%/10%/10% and use 10-fold cross validation for evaluation. |
| Hardware Specification | Yes | For model training, each model is trained for 300 epochs on the Linux platform with an Intel Xeon Gold 6240R CPU and an NVIDIA Tesla V100 SXM2 GPU. |
| Software Dependencies | No | The paper mentions 'implemented in Python' and 'built upon Py Torch' but does not specify version numbers for these software components. |
| Experiment Setup | Yes | For the GCN model (Kipf & Welling, 2017), we adopt three GCN layers with 32 hidden dimensions. We use Re Lu as the activation function and global mean pooling as the readout. A dropout layer with dropout probability p = 0.5 is appended after the GCN layers, and followed by a linear layer with softmax activation to map embeddings to the classification probability. |