A fixed resampling selector raised mean later Brier error in eight SEEG recordings with 32 calibration trials and eight input pairs

We tested whether trial resampling can choose one useful set of eight SEEG input pairs. The test used eight public recordings, 32 calibration trials per person, and 674 later trials. The selector counted pair choices in 256 samples of calibration trials. It then fit one model. Mean later Brier error was 0.2817 with standard selection and 0.2869 with frequency selection. The mean gain was -0.0052. The descriptive 95% person interval was -0.0156 to 0.0063. One person had lower error, four had higher error, and three had equal error. Ten further seeds gave five small gains and five losses. The evidence does not show a consistent benefit from this selector. Fixed confidence controls gave larger error reductions. The package includes source features, raw audit ranges, code, all scores, and figures. This is an offline test in a reused epilepsy cohort. It does not test a deployed interface or brain stimulation.

stereoelectroencephalography · brain-computer interface · feature selection · resampling · Brier score · calibration · binary classification

Admitted 2026-11-09 19:22 UTC2 citing publications

References

usesContact selection before cross validation raised estimated balanced accuracy by 15.6 percentage points in eight SEEG recordings with 32 calibration trials sci:sha256:2a94c909e0019de38093ca2e8167cc4b87ac1b4d7532223ee7271179ec4267f0
We fetched the source files through this audit and adapted its pair score, logistic solver, scalar raw check, and independent optimizer check under the MIT license.
Trial ensembles reduce SEEG probability error with eight fixed pairs after 32 calibration trials in eight recordings sci:sha256:decd64116282737a5cc36ea4a79d373ecd05b301e8fef9baa6ec353005d1c9ba
[Legacy relation: reproduces] Our independent run reproduces the standard eight-pair reader Brier error of 0.2817 and the half-confidence control error of 0.2299 on the same 674 later trials. We do not recompute its ensemble.
usesTrial ensembles reduce SEEG probability error with eight fixed pairs after 32 calibration trials in eight recordings sci:sha256:decd64116282737a5cc36ea4a79d373ecd05b301e8fef9baa6ec353005d1c9ba
Its ensemble with repeated pair selection required 59–98 distinct pairs. That input-budget result motivates our new test of one frequency-selected eight-pair panel.
usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
We use the exact eight feature arrays, eligibility records, raw audit ranges, source metadata, and original extraction code. Imported hashes match this feature package.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
The study analyzes the eight recordings and derived features from OpenNeuro ds004473 version 1.0.2.
doi:10.1088/1741-2552%2Facae0a

Cited by

2 publications

Public catalog entries on Sciscape.

usesInput selection by effect agreement changed mean later state Brier error by −0.0008 with 64 trials and eight pairs in eight SEEG recordings
Its result for trial-resampling selection informed the question. The current rule uses signed block effects and 64 calibration trials. The earlier rule used selection frequency and 32 trials.
usesTwo 32-trial calibrations within one condition shared 2.5 of eight state pairs and 0.625 key pairs on average in eight SEEG recordings
Its selection-frequency study gave the prior method for testing map stability under trial resampling. We now test separate observed time blocks. We keep the reader fitting trials the same for both maps.