← Back to all news
arXiv AI By Jinhua Wu, Xinliang Zhang

Degraded but Not Entirely Ineffective: PE-Based Deformable Graph Neural Networks

Read the original on arXiv AI →

The paper introduces a Position Encoding-Based Deformable Spatial Aggregation Module (PEBDSAM) that uses a deformable mechanism in position space to identify relevant nodes beyond first‑order neighbors, thereby addressing over‑smoothing, over‑compression, limited receptive fields, and noise from heterophilous graphs. Diagnostic experiments revealed that current offsets are ineffective, yet performance still improves, leading to a streamlined version called PEBSAM and a faster variant, PEBSAM‑Speed. The module is plug‑and‑play and can be integrated into GCN, GAT, GIN, and GraphSAGE, achieving strong results on both homophilous and heterophilous datasets.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv Machine Learning
Oct 16

Size-Transferability of Graph Transformers with Convolutional Positional Encodings

The paper investigates Graph Transformers (GTs) that use Graph Neural Network (GNN)-based positional encodings. By analyzing GTs through manifold limit models, the authors establish a theoretical link between GTs and Manifold Neural Networks (MNNs), showing that GTs inherit transferability guarantees from their positional encodings. Experiments on standard graph benchmarks confirm that GTs scale comparably to GNNs, and a real‑world shortest‑path estimation task demonstrates their practical efficiency on large terrains.

By Javier Porras-Valenzuela, Zhiyang Wang, Teresa Shang, Yusu Wang, Alejandro Ribeiro
llmsbenchmarks
More like this →
arXiv Machine Learning
Sep 22

SiST-GNN: Simultaneous Spatial-Temporal Message Passing for Dynamic Graph Representation Learning

SiST‑GNN introduces a simultaneous spatial‑temporal message‑passing framework for dynamic graph neural networks, fusing per‑node temporal embeddings with spatial aggregation in a single operation. By maintaining a recurrent hidden state per node and treating it as a cross‑time edge, the model jointly reasons over topology and evolution. Experiments on link‑prediction and node‑classification benchmarks show significant improvements over prior methods, achieving up to 158% gains in live‑update link prediction and outperforming discrete‑time baselines by 7–23% in dynamic node classification.

By Shubhajit Roy, Anirban Dasgupta
ragbenchmarks
More like this →