Training for equal contact gains within ±1 voltage dB on each shaft gave exact invariance in eight SEEG fits at C = 0.1
We tested bounded gain training on eight reused SEEG records from OpenNeuro ds004473 version 1.0.2. Each reader used three DPSS tapers and 24 log power features. At C = 0.1, training for equal contact gains within plus or minus one voltage dB on each shaft gave exact invariance. All eight bounded fits matched the exact shaft invariant fit. Training transition points ranged from 0.155 to 0.657 voltage dB. The bounded fit reduced ideal worst-buffer loss relative to the standard fit. Mean clean balanced accuracy fell from 73.60% to 70.49%. Lower gain bounds retained some amplitude information, but their later loss advantage over exact invariance was uncertain. The gain must be constant inside each buffer and across frequencies. This offline clinical task test does not establish live BCI performance. The package supplies exact buffers, code, a training threshold tool, a proof, figures, source records, numerical checks, and repeat instructions.
stereoelectroencephalography · brain-computer interface · gain invariance · robust logistic regression · DPSS · signal processing · training threshold
References
5- usesThree DPSS tapers gave lower mean later Brier error than a Hann taper in eight reused SEEG records with fixed bipolar pairs and 500-sample buffers sci:sha256:bd2b70f0665b6319bf4e85827eb9f5e099defe2ec47485c3b3f543374625d480
We used its three-DPSS power code and preserved its exact input and source lineage through the direct input publication.
- usesOne gain rule per shaft gave lower mean later Brier error than one rule per pair in eight SEEG records under an ideal equal contact gain model sci:sha256:eb4fa104a134dfd17677bfafa8ed541c9dd7952676725012543057f513aedf99
We used its exact digital buffers, trial and pair lists, source metadata, range recipe, shaft group rule, and common standardized coefficient penalty.
- usesTraining for simulated fixed 0.8–1.25 pair gains reduced worst loss and mean clean accuracy in eight SEEG readers sci:sha256:ceb56e351b33c4f6524957cb26958ba89042cae64a86998723edb475e2b9e351
We adapted its convex gain fit, exact score radius, and training threshold diagnostic to shaft groups with a matched invariant fit.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
This is the exact CC0 OpenNeuro dataset version used for the eight SEEG records.
- doi:10.1088/1741-2552%2Facae0a
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.