Figure 1: The model of AutoEncoder.