Sharing video blocks raised mean class accuracy by 2.8 points in nine SEEG records with matched window counts

We tested one fixed video class reader on public SEEG data from nine people. Both split methods used 160 fit windows and 160 test windows per person. Mean balanced accuracy was 56.42 percent when sets shared video blocks and 53.65 percent when complete blocks were separate. The mean difference was 2.77 percentage points and was positive in all nine people. Random labels for whole blocks gave 53.05 and 49.87 percent, respectively. Shared blocks can thus support a higher score without class meaning. The fit methods had different numbers of distinct blocks. These data do not show unseen-video, live BCI, emotion, or stimulation performance. The package preserves source provenance, a verified retrieval recipe, all splits and fitted scores, code, checks, and two figures.

stereoelectroencephalography · validation · block splitting · balanced accuracy · brain-computer interface