16.03.2021 Views

Advanced Deep Learning with Keras

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Chapter 5

# save the model after training the generator

# the trained generator can be reloaded for future

# MNIST digit generation

generator.save(model_name + ".h5")

Figure 5.1.4: The sample outputs of WGAN vs. training steps.

WGAN does not suffer mode collapse in all the outputs during training and testing.

WGAN is stable even under network configuration changes. For example,

DCGAN is known to be unstable when batch normalization is inserted before

the ReLU in the discriminator network. The same configuration is stable in WGAN.

[ 141 ]

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!