STYLEGAN Procedure

OPTIMIZATION Statement

OPTIMIZATION <options>;

The OPTIMIZATION statement specifies the optimizer parameters to use in training the style GAN model. You can specify the following options:

ADAM (BETA1=number BETA2=number)

specifies the optimization method to use in training the GAN model. Currently only the Adam method is supported. You must specify the keyword ADAM when specifying the following suboptions:

BETA1=number

specifies the exponential decay rate for the first-moment estimates.

By default, BETA1=0.0.

BETA2=number

specifies the exponential decay rate for the second-moment estimates.

By default, BETA2=0.99.

LEARNINGRATE=number

specifies the learning rate for the optimizer.

By default, LEARNINGRATE=0.002.

MINIBATCHSIZE=number

specifies the number of images in one minibatch. The number must be a positive integer.

By default, MINIBATCHSIZE=32.

NUMITERS=number

specifies the number of minibatches to be used for training the GAN model. The number must be a positive integer.

By default, NUMITERS=10.

Last updated: September 23, 2026