Fig. 17

ROC curves of DeepRAM models [33] with two different data encoding schemes: one-hot encoding (OneHot) and embedding layer (Embed) on three datasets of increasing size. The higher the curve, the better performance with a 45\(^{\circ }\) dashed line to represent random prediction. The splice data has three panels since ROC curves assume binary classification and the splice dataset has three classes (0, 1, 2). Each panel corresponds to prediction one class vs the other two combined. CNN corresponds to the convolutional model, RNN corresponds to the recurrent model and CNN-RNN corresponds to the combined model. Accuracy decreases with data size, and all models display a similar behavior on the different data encoding schemes