Abstract:We present ATLAS (Adaptive Topology-based Learning at Scale for Homophilic and Heterophilic Graphs), a novel graph learning algorithm that addresses two important challenges in graph neural networks (GNNs). First, the accuracy of GNNs degrades when the graph is heterophilic. Second, iterative feature aggregation limits the scalability of GNNs to large graphs. We address these challenges by extracting topological information about graph communities at multiple levels of refinement, concatenating community assignments to the feature vector, and applying multilayer perceptrons (MLPs) to the resulting representation. This provides topological context about nodes and t…
Abstract:We present ATLAS (Adaptive Topology-based Learning at Scale for Homophilic and Heterophilic Graphs), a novel graph learning algorithm that addresses two important challenges in graph neural networks (GNNs). First, the accuracy of GNNs degrades when the graph is heterophilic. Second, iterative feature aggregation limits the scalability of GNNs to large graphs. We address these challenges by extracting topological information about graph communities at multiple levels of refinement, concatenating community assignments to the feature vector, and applying multilayer perceptrons (MLPs) to the resulting representation. This provides topological context about nodes and their neighborhoods without invoking aggregation. Because MLPs are typically more scalable than GNNs, our approach applies to large graphs without the need for sampling. Across a wide set of graphs, ATLAS achieves comparable accuracy to baseline methods, with gains as high as 20 percentage points over GCN for heterophilic graphs with negative structural bias and 11 percentage points over MLP for homophilic graphs. Furthermore, we show how multi-resolution community features systematically modulate performance in both homophilic and heterophilic settings, opening a principled path toward explainable graph learning.
| Comments: | Preprint |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2512.14908 [cs.LG] |
| (or arXiv:2512.14908v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2512.14908 arXiv-issued DOI via DataCite |
Submission history
From: Turja Kundu [view email] [v1] Tue, 16 Dec 2025 20:43:49 UTC (495 KB)