Skip to main content
. 2007 Sep 3;2007:52609. doi: 10.1155/2007/52609

Figure 2.

Figure 2

(a) Classifier error rates for all 10 subjects; regularization parameter and the input time slice were selected to minimize the classification error using 100-fold cross-validation. (b) Control results obtained using the same algorithm on data with randomly scrambled target labels; both plots show average error estimated using 100-fold cross-validation; error bars denote 1-std-wide margin around the estimate.