RNS Limitations
Cross-source consensus on RNS Limitations from 1 sources and 5 claims.
1 sources · 5 claims
Risks & contraindications
Evidence quality
Highlighted claims
- RNS can underperform in structurally extreme settings such as very small dense graphs or highly imbalanced binary-label datasets. — Implicit Regularization of Mini-Batch Training in Graph Neural Networks
- RNS is less clearly suited to architectures with global attention. — Implicit Regularization of Mini-Batch Training in Graph Neural Networks
- The large-scale sampler benchmark uses only three main datasets, leaving some structural explanations tentative. — Implicit Regularization of Mini-Batch Training in Graph Neural Networks
- SGFormer under RNS underperforms full-graph training on OGBN-ARXIV and POKEC, plausibly because all-pairs global attention interactions are truncated. — Implicit Regularization of Mini-Batch Training in Graph Neural Networks
- Rare predictive patterns may be disrupted by random partitioning. — Implicit Regularization of Mini-Batch Training in Graph Neural Networks