Research & experiments
Three design/research deliverables accompany the runtime code. They are documentation and a standalone experiment — none change the verifier, the dashboard, or the runtime data path.
Web UI / CLI verifier
The production path: ingest a BioTrace+ export and emit diagnostics only. No modeling, medical, or mental-state claims. Never imports the experiment or NeuralSet.
Experiment scaffold
Downstream of the verified boundary: windowing, Welch bandpower features, and a baseline classifier with session-aware cross-validation. Its own package and venv; never touches the web request path.
NeuralSet adapter
A gracefully-degrading seam, never a requirement. If the neuralset package is importable it can provide features; otherwise the scaffold falls back to built-in spectral features and reports available=false.
A falsifiable program for multimodal-supervised cognitive-state modeling and sham-controlled closed-loop testing. Fences off consciousness / thought-decoding claims and treats the four-channel montage as a hard constraint.
docs/research/neuromirror-research-thesis.md
An implementation-ready spec mapping NeuroMirror concepts onto NeuralSet's Events / Extractors / Segmenter / Batch abstractions as a separate adapter layer. Keeps NeuralSet / PyTorch out of the web backend.
docs/research/neuralset-integration-spec.md
A standalone, executable baseline on tidy CSV/TSV EEG: validate → normalize markers → conservative preprocess (artifacts flagged, never silently deleted) → event-aligned windowing → Welch bandpower features → baseline classifier with leave-one-session-out cross-validation → metrics.json + plots.
cd experiments/neuralset-experiment python -m venv .venv && source .venv/bin/activate pip install -e ".[dev]" neuralset-scaffold demo --out outputs/demo # synthetic, end-to-end pytest && ruff check src tests # 54 tests, lint