0
$\begingroup$

Building GAN model contains the following steps:

  1. Build generator model, and choose optimization function for this model.

  2. Build discriminator model, and choose optimization function for this model.

  3. Build GAN network:

    3.1 Build model with the architecture of the first 2 models (generator, discriminator)

    3.2 Choose optimization function for this network

There are some issues I'm not sure about them:

  1. It seems that generator model and discriminator model can have different optimization functions, am I right ?
  • What is the purpose of this choice ?
  • Is it realistic and data scientist use this ?
  1. It seems that we can choose different optimization function for GAN model which is not the same as generator or discriminator models.
  • Am I right ?
  • What is the purpose of this choice ?
  • Is it realistic and data scientist use this ?
Improve this question
$\endgroup$
7
  • 1
    $\begingroup$Could you please quote a reference if it is a specialized GAN architecture? In the case of a simple GAN, I believe there would be only two losses i.e. one for Discriminator and one for the GAN.$\endgroup$
    – 10xAI
    CommentedJun 2, 2021 at 4:49
  • $\begingroup$If it is a simple GAN, why would we use 2 different loss functions ?$\endgroup$
    – user3668129
    CommentedJun 2, 2021 at 5:01
  • $\begingroup$Ok. Please clarify one more thing. Do you mean two-loss functions i.e. MSE and cross_entropy Or you mean two-phased training? Since the loss functions are the same so I believe the second question seems more relevant.$\endgroup$
    – 10xAI
    CommentedJun 2, 2021 at 5:10
  • $\begingroup$I mean 2 different loss function (i.e MSE and adam)$\endgroup$
    – user3668129
    CommentedJun 2, 2021 at 5:59
  • 1
    $\begingroup$Adam is not a loss function. It is the optimizer. It would be better if you add the reference code link.$\endgroup$
    – 10xAI
    CommentedJun 3, 2021 at 3:56

0

Reset to default

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.