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

Maximum loss training reduced mean later worst log loss under simulated four sample contact delays in eight SEEG records

We tested whether training for the largest contact delay loss improves a SEEG reader. Public OpenNeuro ds004473 version 1.0.2 supplied eight clinical records. The test used 1,332 early training trials and 674 later test trials. Each fault delayed one physical contact by four samples before bipolar subtraction. Maximum loss training reduced mean later worst group log loss from 0.7230 to 0.6228 nats. The paired difference was -0.1002, with a descriptive 95% person bootstrap interval from -0.1549 to -0.0504. Seven records improved. Equal group weights gave 0.6596 nats. Intact input loss rose by 0.0215 nats. The gain did not hold across all delay sizes. The mean smallest group AUC fell. The package has inputs, code, all fitted readers, predictions, three figures, source records, and repeat instructions. Fresh source checks covered 48 trial ranges. The records and later trials were used in earlier studies. This offline task state test does not establish live BCI performance or stimulation benefit.

stereoelectroencephalography · SEEG · brain-computer interfaces · contact delay · robust logistic regression · group distributionally robust optimization · log loss · offline neural decoding · OpenNeuro

Admitted 0 citing publications

References

8
  1. Full contact delay training reduced later worst Brier error beyond positive weight scaling in eight SEEG state readers at a simulated delay of four samples sci:sha256:c6790a581850c294b64075c303f60bcaee61f0fe912869de758dd27d9d239600

    [Legacy relation: reproduces] The new Newton solver repeated all eight C=1 clean and average delay coefficient vectors and intercepts within 1e-7. Those checks anchor the new rule comparisons.

  2. usesFull contact delay training reduced later worst Brier error beyond positive weight scaling in eight SEEG state readers at a simulated delay of four samples sci:sha256:c6790a581850c294b64075c303f60bcaee61f0fe912869de758dd27d9d239600

    This checked archive supplied the complete numeric input arrays, source records, original models, and retrieval helpers. We kept its trials, time split, pair choices, and scale rule. We fit a new maximum group logistic reader and an equal group mixture control.

  3. usesSimulated one-sample contact delays changed a mean 6.89% of later decisions in eight fixed bipolar SEEG readers sci:sha256:f4e97f0aaab3e15ac3db16e1f58265adf477db9d4f867e5557b55dc2f67ea989

    This study defined delay of a physical contact before bipolar subtraction and supplied the later delay input and retrieval method through the direct parent. We retained all 106 contact cases and four delay doses.

  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

    This benchmark supplied the inherited offline trial eligibility rules, chronological split, native power features, and training pair ranking. We kept those choices fixed for the matched training rule test.

  5. usesTraining with simulated four-sample contact delays reduced later worst-contact Brier error in eight SEEG state readers sci:sha256:c6e0e70775f830b71c0f35a43ff8275c2bc419e8b4a58d86786e49fd9b0555e6

    This study supplied the average contact delay training rule, early delay features, and source recovery code through the direct parent. We used the same four sample training dose, fixed logistic penalty, and test inputs.

  6. usesTraining for fixed contact loss avoided constant models in eight SEEG records under simulated loss with mean replacement. The minimum accuracy gain was uncertain. sci:sha256:6ec5d2a3e933dac2dfe487210ff772b3d0cbe2e97f40a3552404788ac0074764

    This related study used persistent physical contact groups for missing feature training. We use that group distinction to place our contact delay test in context. Our faults act on contact timing before subtraction, with a single common reader and no test contact identity input. We did not copy its hinge loss solver or data transformation.

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

    This exact OpenNeuro dataset version supplied the eight SEEG records and analyzed trials.

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

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.