HIAL: Towards Semantics-Aware Hypergraph Active Learning via Dual-Perspective Information Maximization

Hypergraph Neural Networks (HNNs) model high-order interactions effectively but rely on costly node annotations, motivating Hypergraph Active Learning (HAL). However, many HAL pipelines adapt graph-based querying through clique expansion, which introduces structural bias and can cause ranking collapse, making utilities overly determined by hyperedge cardinalities rather than informative high-order context. We propose HIAL (Hypergraph Influence-based Active Learning), a training-free framework that formulates hypergraph active learning as influence maximization over a high-order context-based weighted pairwise projection of the hypergraph. HIAL employs a High-Order Interaction-aware propagation mechanism that modulates pairwise influence weights using hyperedge cardinality and feature consistency, yielding a scalable linear diffusion process tailored to homophilic hypergraphs. We further combine feature-space coverage and structural reachability into a monotone submodular selection objective, enabling an efficient lazy greedy solver. Experiments on eight benchmarks demonstrate that HIAL consistently outperforms strong baselines across diverse homophilic hypergraph domains.

Many real-world datasets are best described not by simple pairwise links, but by group relationships: a paper may cite several papers, a recipe may contain many ingredients, and a user may review multiple businesses. These group relationships are naturally modeled as hypergraphs. However, training machine learning models on hypergraphs often requires many labeled examples, and obtaining these labels can be expensive. This work studies how to choose a small set of useful nodes to label before training the model. Existing methods often convert each group relationship into many pairwise links, but this can overemphasize large groups and make the selection process biased. We propose HIAL, a training-free method that estimates which nodes are most informative by combining two signals: how well a node can spread useful information through the hypergraph, and how well the selected nodes cover diverse regions of the data. Across eight benchmark datasets, HIAL selects more informative labels, improves model accuracy, and scales to large hypergraphs. This can reduce annotation costs in applications involving scientific papers, social reviews, biological data, and recipes.