sir why we didnt take ground_truth_y also as 0.9 instead of 1
Sir why we didnt take ground_truth_y also as 0.9 instead of 1
The reason is that Label Smoothing only helps in the training of the Discriminator. It makes the discriminator less prone to adversarial examples and reduces the over-fitting of the discriminator.
Now, while training the generator, the discriminator is frozen and we can abandon the label smoothing part. Also, the purpose of the Generator is to fool discriminator into thinking that the fake images are real. The generator wants the discriminator to confidently say that the fake images are real. This is in contrast to what Label Smoothing achieves (makes the discriminator less confident to generalize better).
Hope this helps!
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.