Code and data release for ABR-to-synapse modeling (Liberman and Brad cohorts). Supports Mode A (Zenodo cache → manuscript figures) and Mode B (raw waveforms → full retrain).
cd ABR2synapse
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
# Install Zenodo bundle (see DATA.md)
python scripts/fetch_zenodo_data.py --path ../zenodo-data
# Export figures and tables under figures/paper/ and tables/
python scripts/export_paper.pyVerify the release:
PYTHONPATH=. python scripts/verify_release.py --smoke --mode-a --mode-b --publish| Notebook | Description |
|---|---|
01_cohort_summary.ipynb |
Cohort tables (Tables 1–2, Supp. 3) |
02_stage1_noise_classification.ipynb |
Stage-1 wide noise classifier |
03_stage2_within_cohort.ipynb |
Stage-2 within-cohort RF / XGB |
04_stage2_hp_tuning.ipynb |
Stage-2 hyperparameter search |
05_stage2_synthesis_cv.ipynb |
Stage-2 synthesis cross-validation |
06_univariate_eda.ipynb |
Univariate EDA (Fig. 4) |
07_interpretability_shap.ipynb |
SHAP interpretability (Figs 6–8, Table 6) |
Run from the repo root with PYTHONPATH=. or after pip install -e . if you add packaging later.
- REPRODUCTION.md — step-by-step Mode A and Mode B
- DATA.md — cohort CSVs, Zenodo layout, raw-data policy
- PUBLISH.md — Zenodo + GitHub release checklist (maintainers)
- MANUSCRIPT_2_4.md — §2.4 boilerplate for the paper
| Gate | Command |
|---|---|
| G0 ABRA peak parity | python scripts/abra_peak_parity.py |
| G1 smoke | python scripts/verify_release.py --smoke |
| G2 Mode A manifest | python scripts/verify_release.py --mode-a |
| G3 Mode B SHAP cache | python scripts/verify_release.py --mode-b |
| SP5 publish files | python scripts/verify_release.py --publish |
MIT for code — see LICENSE. Data license is stated on the Zenodo deposit.
See CITATION.cff. Fill Zenodo DOI and GitHub URL before public release.