qrc-statistical-learning

QRC-statistically is an archival research bundle for experiments and manuscript drafts on statistically learnable quantum reservoir computing. It preserves the current paper source, earlier manuscript branches, runnable benchmark drivers, and paper-facing result artifacts in one place.

This is not a polished software release. It is a cleaned archive: some experiment families remain directly runnable from preserved scripts, while others survive only as reports, CSV summaries, and figures.

At a glance

• paper/: current manuscript source, PDF, and figure assets
• experiments/: preserved classical, hybrid, and measurement-aware QRC experiment families
• data/: contextual benchmark tables shipped with the bundle
• legacy_manuscripts/: older theory-first and follow-up drafts

Quick start

1. Create a virtual environment with Python 3.10-3.12 if possible.
2. Install the minimal scientific stack:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -r requirements.txt

3. Run one of the preserved entry points:

python experiments/ads/ads_benchmark.py
python experiments/rads/rads_benchmark.py
python experiments/daqr/daqr_prototype.py
python experiments/scale_qrc/scale_qrc_benchmark.py

make help lists the same setup and run targets in a small wrapper Makefile.

Most preserved drivers write fresh outputs to experiments/<family>/results/. Two older measurement-design demos, experiments/spectra_demo/ and experiments/fused_spectra/, write directly into their own folders, so rerun them in a disposable copy if you want to keep the archival outputs untouched.

Experiment map

• Classical controls: ads, rads
• Hybrid and residual QRC families: daqr, stateful_qrc, dpsr_qrc, scale_qrc
• Measurement-design demos: spectra_demo, fused_spectra
• Results-only paper artifacts: learnability_final, validation

See docs/experiment-families.md for the family-by-family map and docs/running-locally.md for output locations, caveats, and concrete commands.

Reproducibility caveats

• experiments/learnability_final/ contains the final paper-facing tables, figures, and report, but the exact driver for that benchmark is not preserved in this bundle.
• experiments/validation/ preserves a supporting validation note and raw CSVs, not a complete rerunnable pipeline.
• The repository mixes multiple research branches. Shared helpers exist, but the folders are not a single end-to-end package with one canonical entry point.
• I verified the preserved ads, rads, stateful_qrc, qcrx, and dpsr_qrc entry points in this environment. Full reruns remain CPU-heavy, and the measurement-design demos still overwrite archival files in place.