Separate ridge penalties reduced error from a shared penalty for 50 ms offline theta forecasts in seven SEEG rest records
We tested separate penalties for target and helper SEEG histories in seven public rest records from OpenNeuro ds006065 version 1.0.0. Each model used 56 s for fitting and 56 s for tuning. The forecast horizon was 50 ms. The main phase target used a centered 2 s theta filter. Later mean error fell from 62.05 degrees with a shared penalty to 57.90 degrees with separate penalties. Error fell in all seven records. The descriptive 95% person interval for the reduction was 2.93 to 5.39 degrees. Tuning omitted helpers for 10 of 28 targets and used a stronger helper penalty for the other 18. Target-only error was 57.76 degrees. Separate penalties did not improve this mean. A shared search with the same 35 candidate count had 62.12 degrees of error. This exploratory study reuses a known cohort and test interval. It measures offline reference agreement, without a live device or stimulation outcome. The package preserves source hashes, input data, code, models, forecasts, figures, and repeat instructions.
SEEG · theta phase · phase forecasting · ridge regression · separate penalties · intracranial EEG · offline prediction · brain-computer interfaces
References
5- usesAdding three SEEG pairs increased mean 50 ms offline theta phase error after 56 s of ridge fitting in seven rest recordings sci:sha256:37e0a21c84e6afb5f766949c16aad144051ab4e2a2974728bf474e5bad706520
Uses the seven SEEG input files, fit and tuning intervals, later test interval, model code, and phase reference. The target-only and shared scores were reproduced. This study adds separate group penalties and control searches.
- usesLearned filters reduced mean 50 ms offline theta phase error by 12.06 degrees against a calibrated sinusoid in later SEEG rest from seven people sci:sha256:3adc1388ee2f8a7376426deef4c35ed0f92b2d041c8f0962bbb58c3d1b7fb38a
Uses the earlier phase forecast framework and its offline reference definitions, as carried forward by the parent benchmark.
- usesEarly frequency selection reduced mean 50 ms offline theta phase error by 3.55 degrees in seven SEEG rest recordings sci:sha256:5c3ae075960de0774e0fcc176ee07a5e3c9a353d86238f1e49023a23aa4d2bfa
Uses the early separate-contact pair selection code, chosen pairs, snapshot hashes, and source metadata. Fresh full signal downloads reproduced the selected pairs and data hashes.
- usesdoi:10.18112/openneuro.ds006065.v1.0.0
The analysis used seven rest records from this exact OpenNeuro dataset version. The archived source hashes and retrieval records identify the files used; the dataset description lists CC0, and its README states PDDL 1.0.
- doi:10.1038/s41467-025-59417-7
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.