English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Adversarial learning methods have been proposed for a wide range ofapplications, but the training of adversarial models can be notoriouslyunstable. Effectively balancing the performance of the generator anddiscriminator is critical, since a discriminator that achieves very highaccuracy will produce relatively uninformative gradients. In this work, wepropose a simple and general technique to constrain information flow in thediscriminator by means of an information bottleneck. By enforcing a constrainton the mutual information between the observations and the discriminator'sinternal representation, we can effectively modulate the discriminator'saccuracy and maintain useful and informative gradients. We demonstrate that ourproposed variational discriminator bottleneck (VDB) leads to significantimprovements across three distinct application areas for adversarial learningalgorithms. Our primary evaluation studies the applicability of the VDB toimitation learning of dynamic continuous control skills, such as running. Weshow that our method can learn such skills directly from \emph{raw} videodemonstrations, substantially outperforming prior adversarial imitationlearning methods. The VDB can also be combined with adversarial inversereinforcement learning to learn parsimonious reward functions that can betransferred and re-optimized in new settings. Finally, we demonstrate that VDBcan train GANs more effectively for image generation, improving upon a numberof prior stabilization methods.

Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow