Figure 9.

The transfer learning classifier achieved incremental performance gains in classifying real data from an attentional switching task and eventually surpassed the subject-specific benchmark. Mean ± SEM (red, black error bars) for 10 individual (gray) accuracies repeated 25 times when classifying attentional switching. Subject-specific classifier accuracy is indicated by the circular markers (left) and the mean is traced across all pool sizes for comparison (blue dotted line). Chance accuracy is 50% and the data consisted of 38 trials per condition per subject on average.