# dsv-ucopa-reproducibility

**Repository Path**: dirtywave/dsv-ucopa-reproducibility

## Basic Information

- **Project Name**: dsv-ucopa-reproducibility
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-08-25
- **Last Updated**: 2026-08-25

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# DSV-UCOPA reproducibility assets

This repository contains the frozen source code, development protocol,
decision-level predictions, audits, and runtime logs supporting the manuscript
on Decision-level Switch-Value UCOPA (DSV-UCOPA) for cross-bearing ordered
health-stage diagnosis.

## Scientific boundary

The deposited evidence is **performance-observed development evidence**, not an
independent confirmation study. Across 40 physical bearings from UORED-VAFCLS,
XJTU-SY, and PRONOSTIA, DSV-UCOPA improved pooled complete, partial, and
combined macro-F1 relative to UCOPA-v4, retained 65 of 72 beneficial switches,
and reduced harmful switches from 25 to 12. The effect was condition dependent:
XJTU-SY declined by 0.0151 macro-F1, the frozen worst-dataset gate failed, and
leave-one-dataset-out selection fell back to UCOPA-v4 without gain. Therefore,
this release does not support a claim of universal cross-condition superiority.

## Contents

- `src/`: DSV-UCOPA analysis and the minimal local dependency modules.
- `protocols/`: protocol frozen before evaluating the candidate methods.
- `tests/`: leakage and switch-retention unit tests.
- `results/development/`: prediction-level outputs, source-only selection
  audits, bearing effects, and SHA-256 sidecars.
- `logs/`: the single-run console summary and exit code.
- `verify_release.py`: release-integrity and headline-result verifier.

The large public raw bearing datasets are not redistributed. The full analysis
entry point records the required input artifacts and their frozen SHA-256
digests. Re-execution from raw data therefore requires users to acquire
UORED-VAFCLS, XJTU-SY, and PRONOSTIA from their official public sources and
reconstruct the locked inputs described by the protocol.

## Verify the deposit

```bash
python verify_release.py
python -m pytest -q
```

The expected verification result is `PASS`, and the unit-test suite contains
three tests.

## Re-run boundary

The complete development entry point is:

```bash
PYTHONPATH=src python src/analyze_ucopa_alternatives_v4.py --help
```

Its required command-line inputs intentionally include the frozen UORED,
XJTU-SY, and PRONOSTIA summaries/audits. These potentially large upstream
artifacts are not duplicated in this compact release; their required digests
are fixed in `protocols/ucopa_alternatives_v4_development_protocol.json`.

## License

Code is released under the MIT License. Protocols, result tables, audits, and
documentation are released under CC BY 4.0; see `LICENSE-DATA`.