Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Variational Discriminator Bottleneck

Yawei Luo

  • Be the first to comment

  • Be the first to like this

Variational Discriminator Bottleneck

  1. 1. VARIATIONAL DISCRIMINATOR BOTTLENECK: IMPROVING IMITATION LEARNING, INVERSE RL, AND GANS BY CONSTRAINING INFORMATION FLOW Yawei Luo
  2. 2. Notoriously D & G D can always find out the nonessential information from G(z) to make a judgement “fake”. -> Uninformative gradients -> Unstable training! How to force D to focus on essential information of G(z)?
  3. 3. Preliminaries • Mutual Information • Object function in information theoretic view
  4. 4. Preliminaries • Information Bottleneck
  5. 5. Preliminaries
  6. 6. Preliminaries q: decoder E: encoder
  7. 7. Back to GANs
  8. 8. Back to GANs Vanilla GAN: GAN with VIB:
  9. 9. Training I(X, Z) > Ic -> beta ++ I(X, Z) < Ic -> beta --
  10. 10. Experiments - IMITATION LEARNING
  11. 11. Experiments - IMITATION LEARNING
  12. 12. Experiments - INVERSE REINFORCEMENT LEARNING
  13. 13. Experiments – image generation

    Be the first to comment

    Login to see the comments

Yawei Luo

Views

Total views

138

On Slideshare

0

From embeds

0

Number of embeds

0

Actions

Downloads

9

Shares

0

Comments

0

Likes

0

×