Eight added SEEG channels increased mean choice log loss on later trials from six MetaRDK sessions at penalty 10
We tested choice probabilities on later trials from six MetaRDK SEEG sessions, OpenNeuro ds006253 version 1.0.3. Earlier trials set all scales, channel selection, and logistic model coefficients. Six people had 447 test trials. At a fixed penalty of 10, adding eight channels increased mean log loss from 0.577 to 0.617. Each person had the same weight. Error increased in five sessions. The exact two-sided person sign-flip test gave p = 0.0625. The mean gain approached zero at penalty 100. A rule that follows the stimulus reached 72.8% accuracy. The combined model reached 67.4%. These results apply to processed signals and known movement times. They do not establish live BCI performance or a general lack of choice information. The package contains code, source features, labels, model states, probabilities, and three figures. All six public source files were retrieved again. Their rebuilt features matched the parent files byte for byte. A fresh environment reproduced the numeric results within 1e-12.
stereoelectroencephalography · choice decoding · brain-computer interface · logistic regression · MetaRDK
References
3- usesAdding eight SEEG signals to a task model increases offline confidence forecast error on later trials in five of six MetaRDK sessions sci:sha256:f153e0af57732bd1a666747e35778c555a9091f0c980875e17ea5b2460f566ee
We used its six compact MetaRDK feature sets, source metadata, pinned hashes, and extraction and retrieval code. We rebuilt all six source files and tested choice probability forecasts with a new fixed model and target.
- usesdoi:10.18112/openneuro.ds006253.v1.0.3
The six analyzed sessions and processed high-gamma features came from OpenNeuro ds006253 version 1.0.3.
- supportsdoi:10.1038/s41467-025-61744-8
The source paper documents offline normalization, baseline correction, and signal smoothing discussed in the report.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesOffline eight-channel SEEG models gave 0.484 mean choice AUC within visual directions in 8 new MetaRDK participants
We used the original neural-only choice model code and test probabilities obtained through db7c95d3. These enabled a fixed model comparison on new participants.
- usesEight high gamma SEEG means do not reduce mean forecast loss for later task errors in six MetaRDK records
This analysis informed the distinction between initial movement and final response accuracy. We used final task accuracy, retained changes of mind, and did not infer initial choices.
- usesEight high-gamma means increased mean offline forecast loss for later changes of mind in six MetaRDK SEEG sessions under a fixed ridge model
Use the six-person feature data, behavior files, pinned source receipts, and numerical fitting core. The new target is a later recorded change of mind. The parent tested initial choices and excluded change trials.
- usesEqual weights for stimulus and choice cells gave 0.511 mean conditional choice AUC in six public SEEG tests at penalty 10
We reused the six people, choice label code, fixed early and later trial split, and compact SEEG features. We checked ordinary model probabilities against the preserved parent probabilities. We added weighted training and tests within stimulus direction.
- usesTraining on correct trials gave 52.3 percent balanced choice accuracy on later error trials with eight SEEG channels in six MetaRDK sessions
We used its six MetaRDK feature files, source metadata, retrieval code, label rules, channel selection, and logistic solver. The new test fits correct early trials and scores choice on later errors.