Search code examples
tensorflowconv-neural-networktensorboard

What can I infer from Discriminator and Generator loss?

I have trained a GAN and when looking at the loss graphs ,what should I infer, if the discriminator is too strong ?

enter image description here

enter image description here

Solution

  • I believe this problem is known as discriminator's loss "collapse". Maybe adding more layers to generator or adding LeakyRelue activations and batch normalization helps. More details are provided here.