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

Multi-task reinforcement learning endeavors to accomplish a set of differenttasks with a single policy. To enhance data efficiency by sharing parametersacross multiple tasks, a common practice segments the network into distinctmodules and trains a routing network to recombine these modules intotask-specific policies. However, existing routing approaches employ a fixednumber of modules for all tasks, neglecting that tasks with varyingdifficulties commonly require varying amounts of knowledge. This work presentsa Dynamic Depth Routing (D2R) framework, which learns strategic skipping ofcertain intermediate modules, thereby flexibly choosing different numbers ofmodules for each task. Under this framework, we further introduce a ResRoutingmethod to address the issue of disparate routing paths between behavior andtarget policies during off-policy training. In addition, we design an automaticroute-balancing mechanism to encourage continued routing exploration forunmastered tasks without disturbing the routing of mastered ones. We conductextensive experiments on various robotics manipulation tasks in the Meta-Worldbenchmark, where D2R achieves state-of-the-art performance with significantlyimproved learning efficiency.

Not All Tasks Are Equally Difficult: Multi-Task Deep Reinforcement Learning with Dynamic Depth Routing