sciscape Publications
← All publications
cs.LG · Machine Learning

A confidence margin blocked new output errors and withheld most correct outputs in eight SEEG records under an ideal ±3 dB shaft gain model

We tested a gain margin for eight frozen SEEG readers on 975 later task trials. SEEG means stereoelectroencephalography. Each selected shaft could have an independent constant voltage gain within ±3 dB. The model assumes a linear signal chain, fixed labels, and no added noise, clipping, or spectral change. The margin makes every current accepted label agree with a nominal early accepted label. Mean maximum wrong output yield fell from 43.17% to 7.32%. Mean minimum correct output yield fell from 34.35% to 0.75%. All yields use every offered window as the denominator. The accepted-error fraction could still rise. One permitted gain changed it from 7/63 to 6/40 with no new wrong output. The package includes an exact gate tool, all features, source ranges, complete constant-gain audits, figures, and repeat instructions. All 3,133,395 features matched a fresh public source rebuild. The result is an offline test of a reused cohort. It does not establish live control or stimulation performance.

stereoelectroencephalography · brain-computer interfaces · selective classification · gain robustness · abstention · output yield · score intervals

Admitted 1 citing publication

References

7
  1. usesA simulated 6 dB shaft voltage loss raised mean accepted label error by 10.5 points in eight SEEG readers with fixed early confidence gates sci:sha256:c11916a581fda4693fef923ed62c5f6bed20b0fc31fcea1f0823dc4472692661

    This study used its frozen features, models, split rows, early confidence gates, and source ranges. It also used its earlier gain experiment as context.

  2. usesSpectral contrasts gave 61.2% mean balanced accuracy under fixed shaft gains within ±3 dB in eight SEEG recordings sci:sha256:14a3a6432c644bebc872ba12928232e79781ea1a740dff5445b1a5aabf94d042

    This study used its established all-shaft score bound and exact constant-offset sweep.

  3. usesFixed early confidence thresholds reduced mean task state error to 24.5% at 49.6% retention after condition changes in eight offline SEEG tests sci:sha256:67664491c0474529f745f7fca563fcc1e5964887d611b1c9b90a66aa6e227ae6

    This publication supplies the early confidence gates through the direct source package.

  4. usesEight fixed SEEG decoders scored 65.8% mean balanced accuracy for response versus rest windows after task condition changes sci:sha256:27c3a162d5f23c03418feb1337bb50c3dbd6fcf78da3fdc3bdc12da973491084

    This publication supplies the upstream feature extraction and reader fitting through the direct source package.

  5. usesTraining for ±6 dB gain on one shaft raised correct classes with a guarantee in eight SEEG records under a linear gain model sci:sha256:5b0d343335c2da00c225c25a050de434c99e47b5020cc51baebb2b35ef408f94

    This study used its established score interval gate for class stability. The source split differs from this study.

  6. usesdoi:10.18112/openneuro.ds004473.v1.0.2

    The study reuses the eight SEEG records in OpenNeuro ds004473 version 1.0.2; its saved metadata identifies the version-specific DOI and CC0 license.

  7. doi:10.1088/1741-2552%2Facae0a

Cited by

1 publication

Public catalog entries on Sciscape.