REVIEW 9 cited by
Adaptive Universal Generalized PageRank Graph Neural Network
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
In many important graph data processing applications the acquired information includes both node features and observations of the graph topology. Graph neural networks (GNNs) are designed to exploit both sources of evidence but they do not optimally trade-off their utility and integrate them in a manner that is also universal. Here, universality refers to independence on homophily or heterophily graph assumptions. We address these issues by introducing a new Generalized PageRank (GPR) GNN architecture that adaptively learns the GPR weights so as to jointly optimize node feature and topological information extraction, regardless of the extent to which the node labels are homophilic or heterophilic. Learned GPR weights automatically adjust to the node label pattern, irrelevant on the type of initialization, and thereby guarantee excellent learning performance for label patterns that are usually hard to handle. Furthermore, they allow one to avoid feature over-smoothing, a process which renders feature information nondiscriminative, without requiring the network to be shallow. Our accompanying theoretical analysis of the GPR-GNN method is facilitated by novel synthetic benchmark datasets generated by the so-called contextual stochastic block model. We also compare the performance of our GNN architecture with that of several state-of-the-art GNNs on the problem of node-classification, using well-known benchmark homophilic and heterophilic datasets. The results demonstrate that GPR-GNN offers significant performance improvement compared to existing techniques on both synthetic and benchmark data.
Forward citations
Cited by 9 Pith papers
-
Same Graph Cross-Task Transfer in GNNs: Protocols and Predictors
Under a fixed leakage-free protocol, NC→LP transfer reliably helps on homophilic graphs while LP→NC helps mainly when LP is easy and NC is unsaturated; homophily and CoTask Score guide mechanism choice.
-
HeAD-CP: Heterophily-Aware Diffused Conformal Prediction Sets for Graph Neural Networks
Graph conformal prediction with node-wise softmax-derived diffusion coefficients avoids DAPS's heterophily failure and beats it on 8/10 benchmarks in oracle selection.
-
NodeImport: Imbalanced Node Classification with Node Importance Assessment
A closed-form importance score — the gradient alignment between a node and a balanced meta-set — filters labeled, unlabeled, and synthetic nodes, improving GNN balanced accuracy under class imbalance.
-
A Hierarchical Quantized Tokenization Framework for Task-Adaptive Graph Representation Learning
QUIET is a hierarchical RVQ-based graph tokenizer with a learned level-weighting gate; it improves several benchmarks but not consistently against the strongest baselines.
-
Node-as-Agent: Graph Agentic Network
A node-as-agent framework where a frozen LLM plans each node's local and global message passing achieves competitive Cora accuracy without training, but uses per-dataset prompt selection and leaves label-leakage quest...
-
Disentangling Homophily and Heterophily in Multimodal Graph Clustering
DMGC clusters multimodal multi-relational graphs by disentangling homophilic and heterophilic edges and fusing low-pass and high-pass filtered representations in a self-supervised way, reporting SOTA accuracy on six b...
-
ADMP-GNN: Adaptive Depth Message Passing GNN
ADMP-GNN adaptively picks each node's message-passing depth via a centrality-based policy, giving small and inconsistent accuracy gains over fixed-depth GNNs.
-
Flow Matters: Directional and Expressive GNNs for Heterophilic Graphs
A GAT model with polynomial gating (Poly) and a directed variant (Dir-Poly) report strong heterophilic node classification results, with Dir-Poly's largest gain on a single directed dataset.
-
IMPA-HGAE:Intra-Meta-Path Augmented Heterogeneous Graph Autoencoder
IMPA-HGAE masks and reconstructs features and meta-paths in heterogeneous graphs and propagates intermediate meta-path node information, yielding competitive but not uniformly best node-classification results across f...
Discussion (0). Sign in to comment.