Feature Transportation Improves Graph Neural Networks

Authors: Moshe Eliasof, Eldad Haber, Eran Treister

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To demonstrate its efficacy, we evaluate ADRGNN on real-world node classification and spatio-temporal datasets, and show that it improves or offers competitive performance compared to state-of-the-art networks.
Researcher Affiliation Academia 1 Department of Applied Mathematics and Theoretical Physics, University of Cambridge, United Kingdom 2 Department of Computer Science, Ben-Gurion University of the Negev, Israel 3 Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia, Canada me532@cam.ac.uk, ehaber@eoas.ubc.ca, erant@cs.bgu.ac.il
Pseudocode Yes Algorithm 1: Graph Neural Advection-Diffusion-Reaction Layer
Open Source Code No The paper does not explicitly state that the source code for their methodology is made publicly available, nor does it provide a link to a code repository.
Open Datasets Yes We experiment with Cora, Citeseer, and Pubmed datasets. ... Squirrel, Film, and Chameleon from (Rozemberczki, Allen, and Sarkar 2021), as well as the Cornell, Texas and Wisconsin datasets from (Pei et al. 2020)... Chickenpox Hungary, Pedal Me London, and Wikipedia Math datasets from (Rozemberczki et al. 2021), as well as the traffic speed prediction datasets METR-LA (Jagadish et al. 2014) and PEMS-BAY (Chen et al. 2001).
Dataset Splits Yes We use the 10 splits from (Pei et al. 2020) with train/validation/test split ratios of 48%/32%/20%, and report their average accuracy in Table 1.
Hardware Specification No The paper does not explicitly describe the specific hardware used (e.g., GPU/CPU models, memory) to run its experiments in the provided text.
Software Dependencies No The paper mentions using 'Py Torch-Geometric-Temporal' but does not provide specific version numbers for this or other key software components, which are required for a reproducible description.
Experiment Setup No The paper mentions loss functions and refers to external sources for training and testing procedures, but it does not explicitly provide concrete hyperparameter values or detailed system-level training settings in the main text, stating that 'Architectures and training details are provided in the Appendix'.