Figure 4.

The experimental results in the partial agreement setting of the SCIAN dataset: (a,b) Typical classification accuracy and cost curves with the number of epochs during the training on the training and test sets; (c) The confusion matrix on the test set; (d–f) The precision, recall, and F₁-score curves of five classes respectively on the test set; (g) The precision-recall curves of five classes on the test set as well as their micro-averaging precision-recall curve; (h) The receiver operating characteristic (ROC) curves of five classes on the test set as well as their macro and micro-averaging ROC curves.