Bands from four calibration passages per person covered 85 of 100 later sound traces with 117 points in nine SEEG recordings
We tested uncertainty bands for heard sound amplitude read from public SEEG data in OpenNeuro ds004703 version 1.1.0. For each of nine people, two selected early passages fit a frozen reader and four separate passages set the bands. The test used 100 later traces. Each trace had 117 target values from 3.0 to 14.6 seconds. Both band rules had a nominal coverage of 80%. Bands set from single time rows covered 79.59% of rows by an equal-person mean, but covered every target in no test trace. Bands set from the largest error per calibration trace covered 85 of 100 whole test traces. Their equal-person whole-trace coverage was 88.88%. Their mean within-person width ratio against row bands was 2.776. The SEEG reader did not have a clear band-width advantage over a constant reader. The input windows imposed a 400 ms delay. This was an offline test on clinical listening data. The test traces have prior use. Chronological data do not establish the exchangeability assumption needed for the conformal bound. The package preserves compact data, pinned public source recipes, code, plans, models, predictions, tables, and three figures. Fresh checks of all used source ranges passed. Offline numerical replay was exact. Analysis on rebuilt source features preserved every coverage count.
conformal prediction · uncertainty bands · whole-trace coverage · stereoelectroencephalography · brain-computer interface · speech · signal processing
References
5- usesSix calibration passages gave a mean 6.5% lower later squared error than two in nine SEEG sound readers at a 400 ms delay sci:sha256:f5145388b8b3b2413f77551421cbd7d8b1f82e06b2852f358c8cd5d4eb6a0e6a
This study uses the compact SEEG power features, sound targets, source identities, and loading and ridge code from this package. Two early passages fit the reader. Four separate passages set uncertainty bands.
- usesFixed SEEG readers had higher mean sound amplitude correlation at 400 ms than at zero delay on 100 unused later passage records in nine people sci:sha256:3c8b962f91454c5d1b32469fb10a6ecb9534841b65f6b7ab946bc56a062e2126
This study uses the corrected source clock and the fixed set of 100 later test records defined by this work. The test records have prior use and are not a new holdout set.
- usesSelected SEEG pair deviations reduced mean error in log sound amplitude on 100 later passages in nine people with a 400 ms output delay sci:sha256:a2f979743f761bc5638efa44fcc4e1fe5803608c1ad497cd8ed4a35d31ce82af
This study uses the compact adjacent-pair power features and target extraction inherited from this work. It preserves the source notices and checks the features against fresh public source ranges.
- usesdoi:10.18112/openneuro.ds004703.v1.1.0
The analysis uses OpenNeuro ds004703 version 1.1.0 for clinical SEEG recordings, heard-speech stimuli, channel labels, clock offsets, and task records.
- doi:10.1038/s41467-024-44844-9
This is the source study for the conversational-speech experiment and intracranial recordings.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- At an 80% target, limits set with other sounds gave lower mean coverage for voices in 20 iEEG patients and 24 fMRI adults
Related temporal-trace calibration benchmark; our target is a neural response map under sound-family shift.
- usesA ridge-leverage gate showed no clear mean skill gain for later sound estimates from nine SEEG recordings
Used its compact public SEEG and sound arrays, fixed delayed ridge reader, corrected clock, source range recipes, loading code, and inherited notices. Reproduced its nine primary reader predictions exactly.