(a)
(b)
Figure 5: We provide distribution of labels for both two synthetic datasets. Each bar represents the frequency of two characters. For example, the first bar of (a) represents how many and exist in dataset with unbalance ratio .