If the dataset, shuffle, and noise are all kept the same, is it possible for the same GAN to give different Inception Scores each time?
$\begingroup$
$\endgroup$
1
-
$\begingroup$ Welcome to AI stackexchange. If all parts of your application are deterministic, then your inception score should, given the same input, result in the same score. However, there might be ways in which your application is indeterministic, resulting in different scores each time. $\endgroup$– Robin van HoornCommented Jul 5 at 12:45
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
if you fix the seed (all random factors) you should get similar results. but if you only fix the shuffle and noise, it might be not enough. e.g if you train a GAN, then starting point (initialization) of the GAN is not the same always, so the trained GAN will be different at the end.
