Heterogeneous Ensemble Learning for Hierarchical Multi-label Classification (Supplementary Data)

Supplementary data for the paper "Heterogeneous Ensemble Learning for Hierarchical Multi-label Classification" (Machine Learning Journal, 2026, in submission). This repository contains ESM1_ChEBI_v244_datasets - ChEBI datasets used for training DL base learners (built with https://github.com/ChEB-AI/python-chebai, v1.2.0 and https://github.com/ChEB-AI/python-chebai-graph, v1.0.0) ESM2_DL_checkpoints - Trained DL models ESM3_DL_trust_scores - Trust scores for DL models (on both 2- and 3-STAR ChEBI data) ESM4_DL_trust_scores_3star - Trust scores for DL models (only on 3-STAR ChEBI data) ESM5_disjoint_axioms - Disjointness axioms used in the inconsistency resolution algorithm of chebifier (https://github.com/ChEB-AI/python-chebifier, v1.2.1) ESM6_DL5_symbolic_ensemble_predictions - Base learner predictions for the DL5+Symbolic ensemble (optimised for 2- and 3-STAR ChEBI, not aggregated) ESM7_DL6_symbolic_3star_ensemble_predictions - Base learner predictions for the DL6+Symbolic ensemble (optimised for 3-STAR ChEBI, not aggregated) Authors: Simon Flügel [0000-0003-3754-9016] (1), Martin Glauer [0000-0001-6772-1943] (2), Janna Hastings [0000-0002-3469-4923] (3,4), Till Mossakowski [0000-0002-8938-5204] (1), Christopher J. Mungall [0000-0002-6601-2165] (5), Charlotte Tumescheit [0000-0002-7563-5575] (3,4), Fabian Neuhaus [0000-0002-1058-3102] (2), Aditya Ganesh Khedekar [0009-0003-5454-6105] (1,2) (1) Institute for Computer Science, University of Osnabrück, Neuer Graben 29, 49074 Osnabrück, Germany.(2) Institute for Cooperating Systems, Otto von Guericke University Magdeburg, Universitätsplatz 2, 39106 Magdeburg, Germany.(3) Idiap Research Institute, Rue Marconi 19, 1920 Martigny, Switzerland.(4) Swiss Institute of Bioinformatics, Lausanne, Switzerland.(5) Division of Environmental Genomics and Systems Biology, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA.