09.05.2023 Views

pdfcoffee

Create successful ePaper yourself

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

Chapter 6

Figure 1: Visualizing the architecture of a generator

The discriminator network takes in the images (either generated by the generator

or from the real dataset), and the images undergo convolution followed by batch

normalization. At each convolution step the images get downsampled using strides.

The final output of the convolutional layer is flattened and feeds a one-neuron

classifier layer. In the following diagram, you can see the discriminator:

Figure 2: Visualizing the architecture of a discriminator

The generator and the discriminator are combined together to form the DCGAN. The

training follows in the same manner as before; that is, we first train the discriminator

on a mini-batch, then freeze the discriminator and train the generator. The process is

repeated iteratively for a few thousand epochs. The authors found that we get more

stable results with the Adam optimizer and a learning rate of 0.002.

Next, we'll implement a DCGAN for generating handwritten digits.

[ 199 ]

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

Saved successfully!

Ooh no, something went wrong!