Hanfeng Zhai

B. Multiple bubble flow The multiple bubble flow is more complicated to be predicted due to its significant variation in physical variables, especially for small training sets. From predictions of the deep learning frameworks shown in Fig. 10, it can be found that generally BubbleNet(s) performs more accurately than the traditional NN, particularly on the level set function. It is not surprising that DNN fails to estimate the level set function due to its remarkable variation in the training set.