A comprehensive collection of Graph Active Learning (GAL) papers.
Graph Active Learning applies active learning to graph-structured data, strategically selecting valuable graph entities (nodes, edges, or subgraphs) for annotation to maximize GNN performance under limited labeling budgets.
We categorize GAL methods along three orthogonal dimensions that capture the fundamental design choices in graph active learning: how entities are queried (AL mode), the quality of obtained labels (annotation type), and the criteria for entity selection (query strategy).
- Sequential: Query one entity per iteration
- Batch: Query multiple entities per iteration
- One-step: Query all entities in a single iteration
- Accurate: Perfect oracle labels
- Inaccurate: Noisy oracle labels
- Uncertainty (U): Select entities where model is least confident
- Diversity (D): Select entities distinct from labeled ones
- Representative (R): Select prototypical entities covering distributions
- Performance (P): Select entities with greatest model impact
- Single: Use one strategy only
- Pipeline: Apply strategies sequentially
- Hyper-parameter: Combine using fixed/adaptive weights
- Optimization: Learn optimal combination via RL/bandits
- Hybrid: Complex multi-stage combinations
| Paper | AL Mode | Anno. Type | U | D | R | P | Combination | Code | Datasets |
|---|---|---|---|---|---|---|---|---|---|
| [GlobalSIP 2017] [MSD] Active sampling for graph-aware classification | Sequential | Accurate | ✗ | ✗ | ✗ | ✓ | Single | - | Coloncancer, Australian |
| [arXiv 2017] [AGE] Active Learning for Graph Embedding | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hyper-parameter | Code | Citeseer, CORA, Pubmed |
| [IEEE TSP 2018] [-] Data-adaptive Active Sampling for Efficient Graph-Cognizant Classification | Sequential | Accurate | ✗ | ✗ | ✗ | ✓ | Single | - | Synthetic (10x10 grid, LFR); Real-world (CORA, CITESEER, Political-blog, multiple UCI/LibSVM datasets) |
| [IJCAI 2018] [ANRMAB] Active discriminative network representation learning | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Optimization | - | Cora, Citeseer, Pubmed |
| [ICASSP 2018] [AG-SSL] On the Supermodularity of Active Graph-Based Semi-Supervised Learning with Stieltjes Matrix Regularization | Sequential | Accurate | ✗ | ✗ | ✗ | ✓ | Single | - | Karate club network, Dolphin social network |
| [IJCAI 2019] [ActiveHNE] ActiveHNE: Active Heterogeneous Network Embedding | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Optimization | - | DBLP, Cora, MovieLens |
| [arXiv 2019] [FeatProp] Active Learning for Graph Neural Networks via Node Feature Propagation | One-Step | Accurate | ✗ | ✗ | ✓ | ✗ | Single | Code | Cora, Citeseer, Pubmed, CoraFull |
| [SIGKDD 2019] [RRQR] Graph-based semi-supervised & active learning for edge flows | Batch | Accurate | ✗ | ✗ | ✗ | ✓ | Single | Code | Synthetic flows (on Minnesota road network, etc.); Real-world flows (Transportation networks, Last.fm) |
| [IJCAI 2019] [DALAUP] Deep Active Learning for Anchor User Prediction | Batch | Accurate | ✓ | ✗ | ✗ | ✓ | Optimization | Code | Foursquare and Twitter |
| [WWW 2020] [ATNE] Active domain transfer on network embedding | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hybrid | - | Citation Network, Co-author Network |
| [ICML 2020] [GEEM] Active Learning on Attributed Graphs via Graph Cognizant Logistic Regression and Preemptive Query Generation | Single | Accurate | ✓ | ✗ | ✗ | ✓ | Hyper-parameter | - | Cora, Citeseer, Pubmed, Amazon-Photo, Amazon-Computers |
| [NeurIPS 2020] [GPA] Graph Policy Network for Transferable Active Learning on Graphs | Sequential | Accurate | ✓ | ✗ | ✓ | ✗ | Optimization | Code | Reddit, Cora, Citeseer, Pubmed, Coauthor |
| [ACML 2020] [MetAL] Metal: Active semi-supervised learning on graphs via meta-learning | Sequential | Accurate | ✗ | ✗ | ✗ | ✓ | Single | Code | Cora, Citeseer, Pubmed, Coauthor-Physics, Coauthor-CS, Amazon-Computers |
| [TNNLS 2020] [Seal] Seal: Semisupervised adversarial active learning on attributed graphs | Sequential | Accurate | ✗ | ✓ | ✗ | ✗ | Single | - | Citeseer, Cora, Pubmed, DBLP |
| [SIGKDD 2020] [ASGN] ASGN: An active semi-supervised graph neural network for molecular property prediction | Batch | Accurate | ✗ | ✓ | ✗ | ✗ | Single | Code | QM9, OPV |
| [VLDB 2021] [Grain] Grain: Improving Data Efficiency of Graph Neural Networks via Diversified Influence Maximization | Batch | Accurate | ✗ | ✓ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, Pubmed, Reddit, ogbn-papers100M |
| [SIGMOD 2021] [ALG] ALG: Fast and Accurate Active Learning Framework for Graph Convolutional Networks | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, PubMed, Reddit |
| [NeurIPS 2021] [RIM] RIM: Reliable Influence-based Active Learning on Graphs | Batch | Inaccurate | ✗ | ✗ | ✓ | ✗ | Single | Code | Cora, Citeseer, PubMed, Reddit |
| [WWW 2021] [ATTENT] Attent: Active attributed network alignment | Batch | Accurate | ✗ | ✗ | ✓ | ✗ | Single | Code | ACM Citation, DBLP Citation, Douban, Lastfm, Flickr, AMiner |
| [IEEE TRANSACTIONS ON BIG DATA 2022] [ASGNN] Active and semi-supervised graph neural networks for graph classification | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hyper-parameter | - | MUTAG, PTC_MR, COLLAB, BZR_MD, BZR, NCI1, PROTEINS, ER_MD, COX2_MD, DHFR, DHFR_MD, PTC_FR |
| [CIKM 2022] [SMARTQUERY] SmartQuery: An Active Learning Framework for Graph Neural Networks through Hybrid Uncertainty Reduction | Sequential | Accurate | ✗ | ✗ | ✓ | ✓ | Hybrid | - | Cora, Citeseer, Pubmed |
| [AAAI 2022] [BIGENE] Batch Active Learning with Graph Neural Networks via Multi-Agent Deep Reinforcement Learning | Batch | Accurate | ✓ | ✓ | ✓ | ✗ | Optimization | - | Cora, Citeseer, Pubmed, Reddit, Coauthor-CS, Coauthor-Physics |
| [WWW 2022] [ALLIE] ALLIE: Active Learning on Large-scale Imbalanced Graphs | Sequential | Accurate | ✓ | ✗ | ✓ | ✗ | Optimization | - | Cora, Citeseer, Pubmed, a proprietary e-commerce dataset |
| [ICLR 2022] [IGP] Information Gain Propagation: A New Way to Graph Active Learning with Soft Labels | Batch | Accurate | ✗ | ✗ | ✓ | ✓ | Hybrid | Code | Citeseer, Cora, PubMed, Reddit, ogbn-arxiv |
| [SIGKDD 2022] [JuryGCN] JuryGCN: Quantifying Jackknife Uncertainty on Graph Convolutional Networks | Batch | Accurate | ✓ | ✗ | ✗ | ✗ | Single | Code | Cora, Citeseer, Pubmed, Reddit |
| [ECML-PKDD 2022] [LSCALE] LSCALE: Latent Space Clustering-Based Active Learning for Node Classification | Batch | Accurate | ✗ | ✓ | ✓ | ✗ | Pipeline | Code | Cora, Citeseer, Pubmed, Coauthor-CS, Coauthor-Physics |
| [ICDM 2022] [MetRA] Active Heterogeneous Graph Neural Networks with Per-step Meta-Q-Learning | Batch | Accurate | ✓ | ✓ | ✓ | ✗ | Optimization | - | DBLP, ACM, AMiner |
| [NeurIPS Workshop 2022] [AGE+FDS+SDS] Dissimilar Nodes Improve Graph Active Learning | Batch | Accurate | ✓ | ✓ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, Pubmed |
| [LoG 2022] [MMPQ] Jointly Modelling Uncertainty and Diversity for Active Molecular Property Prediction | Batch | Accurate | ✓ | ✓ | ✗ | ✗ | Multiplication | - | Molecular Property Prediction (BACE, BBBP, HIV, SIDER (from MoleculeNet and OGB)) |
| [APWeb-WAIM 2022] [GADAL] GADAL: An Active Learning Framework for Graph Anomaly Detection | Batch | Accurate | ✗ | ✗ | ✓ | ✓ | Hyper-parameter | - | YelpChi, Amazon |
| [ICLR 2023] [SCARCE] Structural Fairness-aware Active Learning for Graph Neural Networks | One-Step | Accurate | ✗ | ✗ | ✓ | ✓ | Optimization | - | Cora, Citeseer, Pubmed, Amazon-Computers, Amazon-Photo, ogbn-arxiv, OGB Products |
| [IJCAI 2023] [GFlowGNN] Generative Flow Networks for Precise Reward-Oriented Active Learning on Graphs | Sequential | Accurate | ✓ | ✗ | ✓ | ✗ | Optimization | - | Cora, Citeseer, Pubmed, 5 Reddit datasets |
| [ECML-PKDD 2023] [DiffusAL] DiffusAL: Coupling Active Learning with Graph Diffusion for Label-Efficient Node Classification | Batch | Accurate | ✓ | ✓ | ✓ | ✗ | Multiplication | Code | Cora, Citeseer, Pubmed, Coauthor-CS, Coauthor-Physics |
| [CIKM 2023] [SAG] Mitigating Semantic Confusion from Hostile Neighborhood for Graph Active Learning | Batch | Accurate | ✗ | ✓ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, Pubmed, Hpay |
| [TKDD 2023] [OWGAL] Open-World Graph Active Learning for Node Classification | Batch | Accurate | ✓ | ✗ | ✗ | ✗ | Hyper-parameter | - | Amazon-Computer, Amazon-Photo, Coauthor-CS, Coauthor-Physics, Arxiv, Reddit, Products |
| [TMLR 2023] [GraphPart] Partition-Based Active Learning for Graph Neural Networks | One-Step | Accurate | ✗ | ✓ | ✓ | ✗ | Pipeline | Code | Citeseer, Cora, Pubmed, Corafull, Ogbn-Arxiv, Co-author CS, Co-author Physics |
| [ICTAI 2023] [ERF-attention] Graph Active Learning at Subgraph Granularity | Batch | Inaccurate | ✓ | ✓ | ✓ | ✗ | Hyper-parameter | - | AMLSim |
| [WWW 2024] [GreedyET] Cost-effective Data Labelling for Graph Neural Networks | Sequential | Accurate | ✗ | ✓ | ✓ | ✗ | Hyper-parameter | Code | Cora, CiteSeer, PubMed, Coauthor Physics, Flickr, Ogbn-arxiv |
| [NeurIPS 2024] [DOCTOR] No Change, No Gain: Empowering Graph Neural Networks with Excepted Model Change Maximization for Active Learning | Sequential | Accurate | ✗ | ✗ | ✗ | ✓ | Single | - | Citeseer, Cora, PubMed, Reddit, ogbn-arxiv |
| [ICDE 2024] [NC-ALG] NC-ALG: Graph-based Active Learning under Noisy Crowd | Batch | Inaccurate | ✗ | ✗ | ✓ | ✗ | Single | Code | Citeseer, Cora, PubMed, Reddit, ogbn-arxiv |
| [UAI 2024] [ALIN] ALIN: An Active Learning Framework for Incomplete Networks | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, Pubmed, Coauthor-CS |
| [arXiv 2024] [MITIGATE] Multitask Active Learning for Graph Anomaly Detection | Batch | Accurate | ✓ | ✓ | ✓ | ✗ | Hybrid | Code | Graph Anomaly Detection (Cora, Citeseer, BlogCatalog, Flickr) |
| [Expert Systems with Applications 2024] [GALMI] Adaptive graph active learning with mutual information via policy learning | Sequential | Accurate | ✗ | ✓ | ✓ | ✗ | Optimization | - | Cora, Citeseer, Pubmed, Chameleon |
| [Communications on Applied Mathematics and Computation 2024] [MC] Model Change Active Learning in Graph-Based Semi-supervised Learning | Batch | Accurate | ✗ | ✗ | ✗ | ✓ | Single | Code | Binary-Clusters (synthetic), MNIST, Salinas A, Urban |
| [SDM 2024] [GALClean] Active Learning for Graphs with Noisy Structures | Batch | Accurate | ✗ | ✓ | ✓ | ✗ | Pipeline | - | Cora, Citeseer, Pubmed, Amazon-photo, Amazon-Computer, Coauthor-CS |
| [IDA 2024] [SPA] A Structural-Clustering Based Active Learning for Graph Neural Networks | Batch | Accurate | ✗ | ✗ | ✓ | ✗ | Pipeline | Code | Citeseer, Pubmed, Corafull, WikiCS, Minesweeper, Tolokers |
| [DMKD 2024] [STAL] Improving graph neural networks by combining active learning with self-training | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Hyper-parameter | Code | Cora, Citeseer, Pubmed, ogbn-arxiv |
| [CIKM 2024] [GraphCBAL] GraphCBAL: Class-Balanced Active Learning for Graph Neural Networks via Reinforcement Learning | Sequential | Accurate | ✓ | ✓ | ✓ | ✗ | Optimization | Code | Cora, Citeseer, Pubmed, Reddit, Coauthor-CS, Coauthor-Physics |
| [TMLR 2025] [LEGO-Learn] LEGO-Learn: Label-Efficient Graph Open-Set Learning | Batch | Accurate | ✓ | ✗ | ✓ | ✗ | Pipeline | - | Cora, AmazonComputers, AmazonPhoto, LastFMAsia |
| [Neural Networks 2025] [DGAL] Disentangled Active Learning on Graphs | Sequential | Accurate | ✓ | ✗ | ✗ | ✗ | Single | - | Cora, Citeseer, Pubmed, DBLP, Reddit, Facebook, Computer, Photo |