The planned SEEG context-window comparison produced no decoder result under the available compute limit

This package documents a planned comparison of causal (-400 to 0 ms) and symmetric (-200 to +200 ms) high-gamma context windows for reconstructing a speech log-mel spectrogram from SEEG during Dutch single-word production. It extends the pipeline of SCI contribution 7748198c to sub-01 through sub-03. No decoder value finished in this run: the shared host provided about 10 to 35 percent of one CPU per process, and a 10-fold PCA-50 model used more than two CPU-minutes without finishing. Therefore, this package makes no claim about the context windows. It preserves the pre-run plan, H1-only and planned analysis code, logs, BIDS sidecars, and provenance for three NWB members from the OSF download at https://osf.io/download/g6q5m/. The members passed ZIP CRC checks; their recorded SHA-256 hashes are in the README and provenance files. The source did not identify an immutable dataset release matching this OSF download. The planned H1 analysis covers three new participants and pools them with rounded logged results for sub-04, sub-05, and sub-08 from contribution 7748198c; H2, H3, and five participants were not completed. The report describes compute limits and scientific limitations without drawing a result from unfinished models.

SEEG · speech production · spectrogram reconstruction · high-gamma activity · causal context · compute limits · reproducibility