Beyond Homophily with Graph Echo State Networks
This addresses the challenge of graph neural network bias in node classification for researchers working with non-homophilous graphs, though it appears incremental as it applies an existing method to a new problem.
The authors tackled the problem of over-smoothing bias in semi-supervised node classification on graphs with varying homophily levels by evaluating Graph Echo State Networks (GESN) for the first time on such tasks, showing they achieve better or comparable accuracy to deep models with architectural variations while being more efficient.
Graph Echo State Networks (GESN) have already demonstrated their efficacy and efficiency in graph classification tasks. However, semi-supervised node classification brought out the problem of over-smoothing in end-to-end trained deep models, which causes a bias towards high homophily graphs. We evaluate for the first time GESN on node classification tasks with different degrees of homophily, analyzing also the impact of the reservoir radius. Our experiments show that reservoir models are able to achieve better or comparable accuracy with respect to fully trained deep models that implement ad hoc variations in the architectural bias, with a gain in terms of efficiency.