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

Offline reinforcement learning -- learning a policy from a batch of data --is known to be hard for general MDPs. These results motivate the need to lookat specific classes of MDPs where offline reinforcement learning might befeasible. In this work, we explore a restricted class of MDPs to obtainguarantees for offline reinforcement learning. The key property, which we callAction Impact Regularity (AIR), is that actions primarily impact a part of thestate (an endogenous component) and have limited impact on the remaining partof the state (an exogenous component). AIR is a strong assumption, but itnonetheless holds in a number of real-world domains including financialmarkets. We discuss algorithms that exploit the AIR property, and provide atheoretical analysis for an algorithm based on Fitted-Q Iteration. Finally, wedemonstrate that the algorithm outperforms existing offline reinforcementlearning algorithms across different data collection policies in simulated andreal world environments where the regularity holds.

Exploiting Action Impact Regularity and Exogenous State Variables for Offline Reinforcement Learning