Why is `use_bias=False` in the GAN tutorial?

TensorFlow
,
1

I am following this tutorial Deep Convolutional Generative Adversarial Network  |  TensorFlow Core While I know what the bias is but I don’t know why it is turned off for the generator in the tutorial and there is no explaination for it. Does it have any performance and/or accuracy impact?

2

I have tried removing the setting use_bias=False and it makes no noticeable difference to me at all… Puzzled.

4

Hi @khteh, In the code, Batch normalization is using. BatchNormalization has its own parameters. So, when we are using conv2d followed by dense or batch normalization then the bias should be fasle.

6

:index_pointing_up:

Any official documentation on your comment?

8

Hi @khteh, I have raised PR for it and it got merged. Please take a look into this for more details. Thanks!