المدربA generative adversarial network (GAN) trains two networks against each other: a generator that maps noise to samples, and a discriminator that tells generated samples from real ones; the generator is trained to make the discriminator fail (Goodfellow and co-authors, 2014). Mode collapse is the failure in which the generator covers only part of the data’s distribution (a few typical patterns, no extremes) because that is enough to fool the current discriminator.
Quantitative Finance · المسرد
ما معنى Generative adversarial network, mode collapse؟
يُعرف أيضًا باسم: generative adversarial network · mode collapse