2026-10-14T03:26:34Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33968662 2026-09-22T19:21:14Z category_146 category_734 category_931 category_811 category_132 portal_87 item_type_3 month_year_09_2026

p–LSGR: A lightweight post–hoc latent–space resampling framework for imbalanced and complex graph distributions Olumayowa Onabanjo (25091542) Gemma Martínez Huerta (25091545) Carlos Francisco Moreno-García (17296120) Marina Díaz Piloñeta (25091548) Joaquín Manuel Villanueva Balsera (25091551) Space Science Biological Sciences not elsewhere classified Information Systems not elsewhere classified Mathematical Sciences not elsewhere classified Infectious Diseases target =' _blank reducing processing time often computationally expensive mitigation strategies provide computationally efficient alternative across diverse benchmarks achieving balanced trade original graph structure graph neural networks complex graph distributions real electrical layouts complex data distributions based resampling strategies findings position p imbalanced graph learning class classification performance p – lsgr space resampling class performance based approaches world settings support research remaining competitive remain sensitive propose post partial robustness lsgr achieves lsgr ), limiting reliability leave post existing imbalance competing approaches class overlap class imbalance class f1 bayesian analysis average 42 additionally release 91 %. 1 </ <p>Graph neural networks (GNNs) are widely used across domains but remain sensitive to class imbalance, class overlap, and complex data distributions, limiting reliability in real-world settings. Existing imbalance-mitigation strategies provide only partial robustness, are often computationally expensive, and leave post-hoc imbalance handling largely unexplored. We propose Post-hoc Latent Space Graph Resampling (p-LSGR), a framework that exploits learned graph representations through latent-space resampling to improve minority-class learning without graph augmentation or modification of the original graph structure. Through data-aware, difficulty-aware, and ensemble-based resampling strategies, p–LSGR improves minority-class classification performance over vanilla GNNs while remaining competitive with augmentation-based approaches, with the ensemble-based variants demonstrating greater stability across encoder types. Across diverse benchmarks, p-LSGR achieves an average 42% improvement in minority-class F1 over vanilla GNNs and a 6% improvement over competing approaches while reducing processing time by up to 91%. Bayesian analysis further indicates that p-LSGR variants maintain strong minority-class performance while achieving balanced trade-offs across Minority F1, Accuracy, and AUPRC. These findings position p-LSGR as a practical and computationally efficient alternative to graph-level augmentation. We additionally release a corpus of constraint-based synthetic graph datasets derived from real electrical layouts to support research on class-imbalanced graph learning at <a target="_blank" href="https://data.mendeley.com/datasets/4vdpxdd7vp/1">https://data.mendeley.com/datasets/4vdpxdd7vp/1</a></p> 2026-09-22T19:21:14Z Dataset Dataset 10.6084/m9.figshare.33968662.v1 https://figshare.com/articles/dataset/p_LSGR_A_lightweight_post_hoc_latent_space_resampling_framework_for_imbalanced_and_complex_graph_distributions/33968662 CC BY 4.0