Poster
Good Semi-supervised Learning That Requires a Bad GAN
Zihang Dai · Zhilin Yang · Fan Yang · William Cohen · Ruslan Salakhutdinov

Wed Dec 6th 06:30 -- 10:30 PM @ Pacific Ballroom #111 #None
in Posters Wed »

Semi-supervised learning methods based on generative adversarial networks (GANs) obtained strong empirical results, but it is not clear 1) how the discriminator benefits from joint training with a generator, and 2) why good semi-supervised classification performance and a good generator cannot be obtained at the same time. Theoretically we show that given the discriminator objective, good semi-supervised learning indeed requires a bad generator, and propose the definition of a preferred generator. Empirically, we derive a novel formulation based on our analysis that substantially improves over feature matching GANs, obtaining state-of-the-art results on multiple benchmark datasets.

Keywords: Semi-Supervised Learning · Adversarial Networks

Author Information

Zihang Dai (Carnegie Mellon University)
Zhilin Yang (Carnegie Mellon University)
Fan Yang (Carnegie Mellon University)
William Cohen (Google AI)
Ruslan Salakhutdinov (Carnegie Mellon University)

More from the Same Authors