Rethinking Graph Super-resolution: Dual Frameworks for Topological Fidelity

2025-11-12

AI Summary

Graph super-resolution (GSR) aims to reconstruct high-resolution (HR) graphs from low-resolution (LR) inputs to reduce costly data acquisition. Existing GNN-based approaches suffer from two key limitations: (i) matrix-based node super-resolution disrupts graph topology and violates permutation invariance; (ii) edge-weight inference relying solely on node representations hinders scalability and expressive power. To address these, we propose two complementary frameworks: Bi-SR and DEFEND. Bi-SR achieves structural-aware node super-resolution via a GNN-agnostic bipartite graph modeling paradigm. DEFEND enables scalable, topology-adaptive edge-weight inference by leveraging dual-graph transformation and standard node-wise GNNs for edge representation learning. Evaluated on real human connectome data, our methods achieve state-of-the-art performance across seven topological metrics. Furthermore, comprehensive experiments on 12 synthetic datasets confirm their robustness and generalizability across diverse graph structures and degradation patterns.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Graph super-resolution, the task of inferring high-resolution (HR) graphs from low-resolution (LR) counterparts, is an underexplored yet crucial research direction that circumvents the need for costly data acquisition. This makes it especially desirable for resource-constrained fields such as the medical domain. While recent GNN-based approaches show promise, they suffer from two key limitations: (1) matrix-based node super-resolution that disregards graph structure and lacks permutation invariance; and (2) reliance on node representations to infer edge weights, which limits scalability and expressivity. In this work, we propose two GNN-agnostic frameworks to address these issues. First, Bi-SR introduces a bipartite graph connecting LR and HR nodes to enable structure-aware node super-resolution that preserves topology and permutation invariance. Second, DEFEND learns edge representations by mapping HR edges to nodes of a dual graph, allowing edge inference via standard node-based GNNs. We evaluate both frameworks on a real-world brain connectome dataset, where they achieve state-of-the-art performance across seven topological measures. To support generalization, we introduce twelve new simulated datasets that capture diverse topologies and LR-HR relationships. These enable comprehensive benchmarking of graph super-resolution methods.

Problem

• Inferring high-resolution graphs from low-resolution counterparts to avoid costly data acquisition
• Addressing matrix-based node super-resolution that disregards graph structure and lacks permutation invariance
• Overcoming reliance on node representations for edge weight inference that limits scalability and expressivity

Innovation

• Bipartite graph connects LR and HR nodes
• Dual graph maps edges to nodes for inference
• GNN-agnostic frameworks preserve topological fidelity