yalishanda/dynamics-needed-mdn

Dynamics Needed — MDN transformer velocity model

Version: 0.1.0

Transformer with a mixture-density-network (MDN) head that predicts a distribution over per-note velocities ("dynamics") for MIDI drum tracks. Trained on the Expanded Groove MIDI Dataset (E-GMD). Part of the Dynamics Needed thesis project.

Intended use

Given a MIDI drum track with flat/undynamic velocities, sample or read out a "best-fitting" velocity per note to restore human-like dynamics. Unlike the LightGBM baseline, this model captures velocity uncertainty.

Training data

E-GMD (Expanded Groove MIDI Dataset), evaluated on the held-out test split.

Metrics (test split)

metric	value
Native NLL	4.473
Discretized NLL	3.132
Deterministic readout — MAE	19.695
Deterministic readout — RMSE	26.641
Sampled — Wasserstein-1	2.948
Sampled — histogram intersection	0.921

Use the deterministic readout for accuracy/ranking; sample from the predicted distribution to restore human-like variance.

Limitations

Published as a versioned research artifact, not a final production model.

• Point-vs-sample trade-off. Sampling restores the dynamic spread that a point estimate flattens, but raises MAE and lowers note-by-note ranking correlation. Pick the readout to match the use case.
• Absolute-loudness generalization gap. On unseen drummers, MAE degrades (~19 → ~28) and the sampled distribution shifts low — a player's overall loudness is not inferable from structure alone. Relative dynamics transfer; absolute level does not.
• Known data artifact. E-GMD's multi-kit rendering remaps pads to different voices per kit, biasing some per-voice results; a single-kit rebuild is a pending fix. See the project's docs/methodology/kit-remapping-artifact.md.
• No listening test yet. Numbers here are offline metrics; perceptual A/B validation is future work.

License

Set to mit by default — change to match the thesis's chosen license before publishing.