What Can We Learn From MIMO Graph Convolutions?

Authors: Andreas Roth, Thomas Liebig

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results confirm that an LMGC can combine the benefits of various methods. Table 1: Results for the universality task. Table 2: Test MAE results on ZINC12k. Table 3: Test accuracy on heterophilic node classification tasks.
Researcher Affiliation Academia Andreas Roth1 , Thomas Liebig1,2 1 Artificial Intelligence, TU Dortmund University 2 Lamarr Insitute for ML and AI, TU Dortmund University EMAIL
Pseudocode No The paper describes methods using mathematical equations and definitions (e.g., Definition 1, Equation 11b) and provides examples for existing methods in the LMGC framework, but does not contain explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Our implementation is available at https://github.com/rothandreas/mimo-graph-convolutions.
Open Datasets Yes We consider the challenging ZINC12k dataset [Sterling and Irwin, 2015]. We consider six heterophilic benchmark datasets for node classification: Texas, Cornell, Wisconsin, Film, Chameleon, and Squirrel.
Dataset Splits Yes We sample a random undirected and connected Erd os R enyi graph [Erd os and R enyi, 1959] with n = 16 nodes and an edge probability of p = 10%. We use the ten splits into train, validation, and test sets proposed by Pei et al. [2020].
Hardware Specification Yes All experiments are run on an H100 GPU.
Software Dependencies No The paper mentions using the Adam optimizer and integrating models into existing frameworks like Graph GPS, Long Range Graph Benchmark, and an implementation from Rusch et al. [2023], but it does not specify version numbers for these or other software libraries (e.g., Python, PyTorch, CUDA).
Experiment Setup Yes The learning rate is tuned in {0.03, 0.01, 0.003}. We optimize the number of layers in {6, 8, 10} and the learning rate in {0.001, 0.0003, 0.0001} using a grid search. Each model utilizes at most 100 000 parameters to ensure fairness. For each method, we tune the learning rate in {0.01, 0.003, 0.001} and dropout ratio in {0.0, 0.25, 0.5} using a grid search.