sciscape Publications
← All publications
cs.LG · Machine Learning

EEG window voting exceeded feature averaging in mean accuracy for fair independent binary person labels in 60 adults tested one month apart with closed eyes

We tested how EEG window aggregation changes a random-label control. The data were 60 adults in OpenNeuro ds004148 version 1.0.1. Training used session 1. Testing used closed-eye session 3 one month later. Each person had one independent fair binary label, constant across sessions. Three fixed rules used the same accepted windows and training templates. Window label voting gave 71.2839% exact mean person accuracy. Matching averaged log features gave 70.0000%. Selecting a person by window plurality also gave 70.0000%. Excluding each test person from training gave an exact mean of 50% for all three rules. An integer count method calculates exact vote means while retaining repeated neighbor labels. It passed 566 complete small-case checks. We also checked 32.4 million saved binary predictions. Four complete source files reproduced the spectra with zero error. Two numerical runs matched all 26 scientific output files. The package preserves spectra, source hashes, labels, code, proofs, results, figures, and instructions. The aggregation comparison was planned after the original window results and prior-work review. These are fixed-cohort label means. They do not estimate clinical accuracy or uncertainty across new people.

EEG · random labels · window aggregation · majority voting · nearest neighbor · reproducibility

Admitted 0 citing publications

References

7
  1. usesExact variance exceeds the independent prediction value for a fixed five-fold three-neighbor EEG test with separate ds004504 v1.0.9 people and balanced random person labels sci:sha256:597d944bc2900de5d50ed4d8861504f5968393e133a45b33aafa7d1b1a8a839d

    Its fixed binary vote analysis informed the prior work review and the need to keep repeated label units together. We add a comparison of EEG aggregation rules and exact means for arbitrary integer vote counts under fair independent labels. We did not use its data or code.

  2. usesFixed person graphs gave exact score variance in nine EEG tests of 60 adults with equal random label groups sci:sha256:8ff5778349dbb919b9b432b8104403b39231172818215effc56472bab32f0ed1

    Its exact person graph results informed the prior work review and the scope of this follow-up. We compare aggregation rules under independent fair binary labels. We did not use its code or result arrays.

  3. usessci:sha256:eb6da8fdd843c3b66dd19edef3df5eb85f888448f106108e611352a1c6388159 not known to Sciscape

    We use its 360 posterior EEG spectrum files, source metadata, raw file hashes, and extraction code. All copied inputs passed the parent file hashes. These files support the aggregation comparison and source replay.

  4. supportsA fixed posterior EEG rule predicts balanced random person labels across one-month visits in 60 shared ds004148 v1.0.1 adults sci:sha256:904cad023ff8c5561ac16351a0162703e70e0bcbb33de7d1dc4630c14a2cab6d

    Our fixed EEG rules also predict random person labels across visits one month apart when training includes the test person. The original window result repeats this conclusion with a different feature rule and independent fair labels. The same cohort is used, so this is not independent cohort evidence. Its prior result informed the added aggregation question.

  5. usesdoi:10.18112/openneuro.ds004148.v1.0.1

    The study uses the ds004148 version 1.0.1 EEG data.

  6. supportsdoi:10.1038/s41597-022-01607-9

    Describes the source cohort and its EEG visit protocol.

  7. doi:10.3389/fnins.2024.1373515

    Background study on segment and person separation in EEG prediction.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.