A reader mixture reduced maximum Brier error compared with a hard route in eight SEEG recordings under uniform simulated shaft loss and 0 to 20 percent wrong flags
We tested a probability mixture in eight public motor-task SEEG recordings from OpenNeuro ds004473 v1.0.2. We simulated loss of one electrode shaft. True shafts had equal weight. Wrong flags named each other shaft with equal probability. Flag errors were independent of signal and class. We used 1068 trials to fit readers, 256 separate trials to set weights, and 674 later trials to test them. The mean of the maximum person Brier errors over 0 to 20 percent wrong flags was 0.20515 for the mixture and 0.20736 for the hard route. The paired change was -0.00221 (95 percent person-bootstrap interval -0.00357 to -0.00097; exact sign-flip p=0.01562). Seven recordings improved. One was unchanged. A fixed 50:50 mixture gave similar error. The package includes an exact calibration rule, proof, source features, models, predictions, figures, and retrieval code. The test trials were used in earlier studies. This result concerns simulated loss. It does not establish live BCI or stimulation performance.
stereoelectroencephalography · brain-computer interfaces · Brier score · probability mixture · calibration · simulated channel loss
References
3- 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 used the exact full-pair source power features, trial eligibility rules, source metadata, and range retrieval code. We selected eight pairs again from a smaller fit block, set mixture weights on a separate calibration block, and tested the original later trial block.
- usesStored recovery readers gave no clear mean Brier error gain with 10 percent wrong shaft flags in eight offline SEEG recordings sci:sha256:3afae3e6eef0a68bf17ed8ff74f016a7a8c62bb5c96cdde8bd6cdabbe33a0586
We used the simulated shaft loss definition, shaft prefix mapping, complete and recovery reader design, and Newton logistic solver. We added a convex probability mixture and exact calibration over a wrong-flag rate interval.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
We used the version 1.0.2 recordings and cite the exact dataset release used in this study.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.