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

As reinforcement learning agents are tasked with solving more challenging anddiverse tasks, the ability to incorporate prior knowledge into the learningsystem and to exploit reusable structure in solution space is likely to becomeincreasingly important. The KL-regularized expected reward objectiveconstitutes one possible tool to this end. It introduces an additionalcomponent, a default or prior behavior, which can be learned alongside thepolicy and as such partially transforms the reinforcement learning problem intoone of behavior modelling. In this work we consider the implications of thisframework in cases where both the policy and default behavior are augmentedwith latent variables. We discuss how the resulting hierarchical structures canbe used to implement different inductive biases and how their modularity canbenefit transfer. Empirically we find that they can lead to faster learning andtransfer on a range of continuous control tasks.

Exploiting Hierarchy for Learning and Transfer in KL-regularized RL