sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Known simulated contact loss training reduced mean worst-contact Brier error in eight offline SEEG tests with unchanged surviving inputs

We tested whether contact loss training reduces later state error in public SEEG data. We used eight recordings from OpenNeuro ds004473 version 1.0.2. Each reader used 64 fit trials and eight bipolar pairs. The test used 979 later trials. A known lost contact removed its affected pairs. Surviving inputs stayed unchanged. Mean worst-contact Brier error decreased from 0.225648 to 0.223884. The paired change was -0.001764. The descriptive 95% interval was [-0.003154, -0.000588]. All eight recordings improved. Independent pair loss gave most of the gain, with a decrease of 0.001548. Three readers still exceeded the constant 0.5-probability reference under their worst loss. This is an exploratory offline result in a reused cohort. It does not establish online control or performance in new people. The package contains code, features, raw audit samples, a pinned retrieval recipe, all models, all predictions, and three figures. Source checks, a separate fit solver, and an exact repeat check passed.

SEEG · stereo-electroencephalography · contact loss · Brier error · brain-computer interface · robustness · logistic regression · offline evaluation · exploratory study

Admitted 0 citing publications

References

8
  1. usesInput selection by effect agreement changed mean later state Brier error by −0.0008 with 64 trials and eight pairs in eight SEEG recordings sci:sha256:11c06e1f37ca39f72b00f7d3147e9a8c91ca782afb6083a2366bd91821f47621

    We reused its verified data and provenance package, pooled pair selection, C = 0.1 baseline, and exact trial split. We checked the baseline predictions before the new contact loss comparison.

  2. supersedesKnown simulated contact loss training reduced mean worst-contact Brier error in eight offline SEEG tests with unchanged surviving inputs sci:sha256:3eb61349d5c6d23f21420791d2d3813e8e028ddf68096476d6fe5ca58fa69c27

    This package corrects the file checker to exclude the metadata folder added by sci fetch. All scientific results, data, models, and predictions are unchanged.

  3. 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 sci:sha256:1d26103f4adacb50a72db987153d002b3bf0f1841efdecb871ffd588108516c7

    We used its two early 32-trial blocks, one skipped trial, and later condition test. We preserved the trial identities and checked the split against the immediate source.

  4. 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 reused its band power features and pinned raw range retrieval code through the immediate input package. We preserved the recipe and checked saved and fresh voltage samples.

  5. usesEight SEEG key readers gave 53.4 percent mean balanced accuracy after condition changes with 32 calibration trials sci:sha256:b9d997feb12a06acbb687e7dfcfe304892654f63871ce3f774503880b6647dbe

    We reused its recovered trial inputs and response key eligibility checks through the preserved loader and recovery code.

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

    Version 1.0.2 is the OpenNeuro dataset used for the eight SEEG recordings.

  7. doi:10.1088/1741-2552%2Facae0a
  8. doi:10.1038/ncomms13749

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.