Training for the largest loss gave higher mean maximum Brier error after temporal tuning in eight SEEG records with rectangular power features and one known absent pair
We tested one SEEG probability reader with one known absent bipolar output. Eight records supplied 2014 trials. Models used rectangular band powers, fixed pairs, and temporal tuning. Training minimized either the mean or the largest logistic loss across nine masks. In 674 later trials, the mean maximum Brier errors were 0.192010 for mean loss and 0.193902 for largest loss. The paired difference was +0.001892. Its 95% person bootstrap interval was [0.000288, 0.003587]. Error was higher in seven records. The two-sided sign-flip test gave p = 0.08594. A common penalty reduced the difference to +0.000175. Hann powers changed it to -0.001446. Both check intervals included zero. The evidence does not show a general benefit from either rule. The cohort and test trials have prior use. Masks simulate absent outputs, without physical contact failure. The package preserves signal samples, source receipts, code, models, scores, three figures, and reproduction steps. A complete repeat matched 23 result files. Independent bounds checked all 104 largest-loss fits.
stereoelectroencephalography · brain-computer interface · robust optimization · Brier score · feature masking · probability calibration · temporal validation
References
5- usesTraining for one known feature group loss increased mean worst-case skill and decreased mean clean skill in nine SEEG sound readers sci:sha256:b4bc3665fe36a489d98947ecc3dda47437984fd090bd9b7123588dca7bd43fbe
We inspected its known group loss study to define the method comparison. Its mean mask training for sound prediction provided design context for testing mean versus largest logistic loss. We did not reuse its data, code, or models.
- usesOne shared gain constraint gave lower mean Brier error than eight pair gain constraints in eight reused SEEG records under a uniform gain model sci:sha256:c23a39e2554b405e4ae3bc346d7725666cbf215170a730abf032b586a85f3661
This publication supplied the exact eight waveform files, selected bipolar pairs, trial eligibility, source metadata, and outer temporal split. We used these fixed inputs for a new comparison of mean and largest logistic loss. We did not reuse its fitted readers or gain comparison.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
This publication defined the inherited fixed pair lists and task windows. The current study keeps those selections and their offline eligibility limits. The exact selection records are preserved in the package.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis used the eight SEEG records and fixed signal samples from OpenNeuro ds004473 version 1.0.2.
- doi:10.1088/1741-2552%2Facae0a
Source study describing the forced two-choice SEEG task and dataset context.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.