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

We study the design of sample-efficient algorithms for reinforcement learningin the presence of rich, high-dimensional observations, formalized via theBlock MDP problem. Existing algorithms suffer from either 1) computationalintractability, 2) strong statistical assumptions that are not necessarilysatisfied in practice, or 3) suboptimal sample complexity. We address theseissues by providing the first computationally efficient algorithm that attainsrate-optimal sample complexity with respect to the desired accuracy level, withminimal statistical assumptions. Our algorithm, MusIK, combines systematicexploration with representation learning based on multi-step inversekinematics, a learning objective in which the aim is to predict the learner'sown action from the current observation and observations in the (potentiallydistant) future. MusIK is simple and flexible, and can efficiently takeadvantage of general-purpose function approximation. Our analysis leveragesseveral new techniques tailored to non-optimistic exploration algorithms, whichwe anticipate will find broader use.

Representation Learning with Multi-Step Inverse Kinematics: An Efficient and Optimal Approach to Rich-Observation RL