Get Started
Home
Topics
Search
Library
Research questionHow can graph neural networks capture long-range interactions without oversmoothing or oversquashing on diverse large graphs?Recursive neighborhood aggregation can blur node representations and compress distant signals as network depth increases. Existing approaches for detecting long-range interactions may also be expensive on large graphs or transfer poorly across graph structures.
AI
Evaluation & Benchmarks
Machine Learning
Neural and Evolutionary Computing
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.DeltaGNN: Graph Neural Network with Information Flow ControlThe source proposes and evaluates DeltaGNN with an information-flow control mechanism for semi-supervised node classification. Its reported evidence covers 10 real-world datasets with varied sizes, topologies, densities, and homophilic ratios, alongside theoretical support and claimed linear computational and memory overhead.research paper · Sep 4, 2026
Related questions
How can reservoir-based graph neural networks aggregate multi-hop information without over-smoothing?How can node-level over-smoothing diagnostics guide virtual-node selection to alleviate over-squashing in GNNs?How can a single graph-learning model handle text-, image-, and multimodal-attributed graphs?How can graph pooling preserve global-to-local topology while adapting granularity for graph classification?