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

While the rapid progress of deep learning fuels end-to-end reinforcementlearning (RL), direct application, especially in high-dimensional space likerobotic scenarios still suffers from low sample efficiency. Therefore StateRepresentation Learning (SRL) is proposed to specifically learn to encodetask-relevant features from complex sensory data into low-dimensional states.However, the pervasive implementation of SRL is usually conducted by adecoupling strategy in which the observation-state mapping is learnedseparately, which is prone to over-fit. To handle such problem, we summarizethe state-of-the-art (SOTA) SRL sub-tasks in previous works and present a newalgorithm called Policy Optimization via Abstract Representation whichintegrates SRL into the policy optimization phase. Firstly, We engage RL lossto assist in updating SRL model so that the states can evolve to meet thedemand of RL and maintain a good physical interpretation. Secondly, weintroduce a dynamic loss weighting mechanism so that both models canefficiently adapt to each other. Thirdly, we introduce a new SRL prior calleddomain resemblance to leverage expert demonstration to improve SRLinterpretations. Finally, we provide a real-time access of state graph tomonitor the course of learning. Experiments indicate that POAR significantlyoutperforms SOTA RL algorithms and decoupling SRL strategies in terms of sampleefficiency and final rewards. We empirically verify POAR to efficiently handletasks in high dimensions and facilitate training real-life robots directly fromscratch.

POAR: Efficient Policy Optimization via Online Abstract State Representation Learning