We have trained a GAN and now would apply to new dataset that it hasn’t seen before. Since the role of discriminator was to help the generator, and it was achieved with training data, for new data, do we need the discriminator or using the generator alone is good enough.
My thinking is that the discriminator would be needed if the dataset changes, but I am throwing this open to the community to get their inputs. I could be incorrect on my line of thinking.