Skip to main content
. 2015 Oct 12;5(12):e00391. doi: 10.1002/brb3.391

Figure 5.

Figure 5

ROC curves for the supervised model and semi‐supervised model. The models were trained on the feature set extracted by BoW21 from the contrast speech versus silence. The parameter r for the semi‐supervised model was fixed to be 0.6. The ROC curves and AUCs were derived based on the predicted values on the testing samples with the LOOCV approach.