search

bbci / bbci_public

MATLAB toolbox for Brain-Computer Interfacing (BCI)
MIT License
148 stars 104 forks source link

Question: Epoch averaging #199

Closed okbalefthanded closed 7 years ago

okbalefthanded commented 7 years ago

Hi,

following the ERPs classification demo and checking the number of epochs used for the stratified 10-fold cross-validation I assume that the training and testing is done on the basis of single-trial i.e. only 1 repetition for each stimulus is taken.

My question is, how epoch averaging is implemented (if it's already programmed) in the toolbox ?

BenjaminBlankertz commented 7 years ago

I would suggest not to average the ERPs, but to use single-trials for classification. The averaging can rather be done afterwards on the classifier output. (However, the difference in performance between averaging epochs and averaging classifier outputs is not large.) Averaging classifier outputs is not implemented in the public toolbox for validation. It is just implemented in the online system (function bbci_control_ERPSpeller.m).

okbalefthanded commented 7 years ago

Thanks for the reply and suggestion.