About the GAN and else

[tontyoutoure]

Hi, when I tried to train the MAP-NN network with my own dataset, the generative error term gen_cost keot increasing, and then converge to a value. Is this suppose to be happen? The mse_cost term and edge_cost term did decrease and converge though.