Topology-Aware Active Learning on Graphs

2025/10/29 by Harris Hardiman-Mostow, Hardiman-Mostow, Harris, Jack Mauro +5 · 1 voice
Computer Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.LG

paper · pdf · doi:10.48550/arxiv.2510.25892

arxiv published 2025/10/29 · arxiv updated 2026/04/15

Abstract

We propose a graph-topological approach to active learning that directly targets the core challenge of exploration versus exploitation under scarce label budgets. To guide exploration, we introduce a coreset construction algorithm based on Balanced Forman Curvature (BFC), which selects representative initial labels that reflect the graph's cluster structure. This method includes a data-driven stopping criterion that signals when the graph has been sufficiently explored. We further use BFC to dynamically trigger the shift from exploration to exploitation within active learning routines, replacing hand-tuned heuristics. To improve exploitation, we introduce a localized graph rewiring strategy that efficiently incorporates multiscale information around labeled nodes, enhancing label propagation while preserving sparsity. Experiments on benchmark classification tasks show that our methods consistently outperform existing graph-based semi-supervised baselines at low label rates.

Citations

• Higher-Order Regularization Learning on Hypergraphs
• Analysis of Semi-Supervised Learning on Hypergraphs
• Graph Pooling via Ricci Flow
• Curvature-based Clustering on Graphs
• Consistency of Fractional Graph-Laplacian Regularization in Semi-Supervised Learning with Finite Labels
• Heterophily-Aware Graph Attention Network
• Coverage-centric Coreset Selection for High Pruning Rates
• Poisson Reweighted Laplacian Uncertainty Sampling for Graph-based Active Learning
• FoSR: First-order spectral rewiring for addressing oversquashing in GNNs
• Graph-based Active Learning for Semi-supervised Classification of SAR Data
• Understanding over-squashing and bottlenecks on graphs via curvature
• Multiscale Laplacian Learning
• Is Homophily a Necessity for Graph Neural Networks?
• PTN: A Poisson Transfer Network for Semi-supervised Few-shot Learning
• Cautious Active Clustering
• Poisson Learning: Graph Based Semi-Supervised Learning At Very Low Label Rates
• A Simple Framework for Contrastive Learning of Visual Representations
• Spectral Analysis Of Weighted Laplacians Arising In Data Clustering
• Graph Interpolating Activation Improves Both Natural and Robust Accuracies in Data-Efficient Deep Learning
• Label Propagation for Deep Semi-supervised Learning
• Properly-weighted graph Laplacian for semi-supervised learning
• Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learning
• Large Data and Zero Noise Limits of Graph-Based Semi-Supervised Learning Algorithms
• Consistency of Lipschitz learning with infinite unlabeled data and finite labeled data
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Analysis of p-Laplacian Regularization in Semi-Supervised Learning
• A variational approach to the consistency of spectral clustering
• A tutorial on spectral clustering
• A Comprehensive Survey on Graph Neural Networks

Discussions

• Topology-Aware Active Learning on Graphs arxiv.org/abs/2510.25892 Label coreset construction by Balanced Forman Curvature* * introduced by Topping (ICLR 2022) arxiv.org/abs/2111.14522 [bsky, 0 points, 0 comments]

Related

• When Contrastive Learning Meets Active Learning: A Novel Graph Active Learning Paradigm with Self-Supervision
• LSCALE: Latent Space Clustering-Based Active Learning for Node Classification
• Active Learning for Graphs with Noisy Structures
• A-Optimal Active Learning
• Active Learning for Node Classification in Assortative and Disassortative Networks