Unrolled Generative Adversarial Networks
Luke Metz
and
Ben Poole
and
David Pfau
and
Jascha Sohl-Dickstein
arXiv e-Print archive - 2016 via Local arXiv
Keywords:
cs.LG, stat.ML
First published: 2016/11/07 (7 years ago) Abstract: We introduce a method to stabilize Generative Adversarial Networks (GANs) by
defining the generator objective with respect to an unrolled optimization of
the discriminator. This allows training to be adjusted between using the
optimal discriminator in the generator's objective, which is ideal but
infeasible in practice, and using the current value of the discriminator, which
is often unstable and leads to poor solutions. We show how this technique
solves the common problem of mode collapse, stabilizes training of GANs with
complex recurrent generators, and increases diversity and coverage of the data
distribution by the generator.