Normal probability averaging reduced mean Brier error after simulated loss of one contact in eight SEEG readers with 5% covariance shrinkage
We tested probabilities after the known loss of one physical contact in eight public SEEG recordings. The source was OpenNeuro ds004473, version 1.0.2. The two labels were rest and the time before a cued response. We fitted eight bipolar pairs per person on early trials. Separate early trials set a confidence scale. We tested 107 contact cases on 674 later trials. Normal averaging used a conditional normal feature model with 5% covariance shrinkage. Mean Brier error fell from 0.20013 for conditional mean filling to 0.19838 for normal averaging. Seven of eight people improved. The mean change was -0.00175, with a descriptive 95% person bootstrap interval from -0.00285 to -0.00088. Early confidence scaling gave lower mean error, 0.18990. Normal averaging and scaling preserved the conditional-mean decisions and within-contact ranks. The result concerns offline task-state probabilities under ideal known feature loss. These data were used in earlier work. The package contains features, source checks, code, fixed models, predictions, figures, and reproduction instructions.
stereoelectroencephalography · SEEG · brain-computer interface · probability estimation · Brier score · missing features · contact loss · calibration · logistic regression
References
3- usesLoss training raised mean accuracy by 0.34 points under simulated contact loss in eight SEEG decoders sci:sha256:884a92921286ae95392544128af6a4f343937c3b760e03bce164eec4599853d4
We reused its complete candidate features, retained trials, physical contact map, source files, and range receipts. We selected new pairs and fitted new readers on a smaller early partition. Its later trials were reused, so this is an exploratory analysis.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
Its public SEEG extraction and eight-pair design supplied the source feature method. We preserved that extraction and tested new probability methods under known physical contact loss.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analyzed SEEG recordings and physical contact identities came from OpenNeuro ds004473, version 1.0.2.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesContact mask training gave no clear mean Brier error gain after two simulated contact losses in eight SEEG recordings
Its one-contact failure comparison informed the new two-contact question and the need to remove every bipolar pair that contains a failed physical contact. We did not reuse its fitted readers or predictions.
- usesLinear feature replacement increased mean sound error after simulated group loss in ten SEEG listeners with eight pairs and 400 ms delay
Uses its conditional normal treatment of known missing inputs as method context. The present study tests linear point replacement for sound amplitude, with a fixed reader.