Figure 4
From: Performance of quantum kernel on initial learning process
The plotting onto ROC space. FPR and TPR show false positive rate and True positive rate. FPR is \(\text{FP}/(\text{FP} + \text{TN})\) on confusion matrix. TPR is \(\text{TP}/(\text{TP} + \text{FN})\), which is the same as Recall. Open diamond and square shape stand for training size of 8 and 200 on heart disease dataset, 6 and 60 on Iris_2 dataset