Reproducibility Archive for "Predicting DEAM Valence and Arousal Annotations from Handcrafted Audio Features — Nonlinear Modeling and Post-Hoc Analysis"

Reproducibility materials accompanying the manuscript “Predicting DEAM Valence and Arousal Annotations from Handcrafted Audio Features — Nonlinear Modeling and Post-Hoc Analysis.” The archive contains the derived 122-feature matrix, analysis scripts, fixed split assignments, canonical feature order, frozen hyperparameters/model manifest, saved development out-of-fold and frozen-test predictions, and derived result tables. It does not redistribute DEAM audio or the original raw/per-rater DEAM annotation files. DEAM-derived fields remain subject to the dataset’s original non-commercial license and attribution requirements.