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

Reinforcement learning (RL) can enable sequential decision-making in complexand high-dimensional environments if the acquisition of a new state-action pairis efficient, i.e., when interaction with the environment is inexpensive.However, there are a myriad of real-world applications in which a high numberof interactions are infeasible. In these environments, transfer RL algorithms,which can be used for the transfer of knowledge from one or multiple sourceenvironments to a target environment, have been shown to increase learningspeed and improve initial and asymptotic performance. However, most existingtransfer RL algorithms are on-policy and sample inefficient, and often requireheuristic choices in algorithm design. This paper proposes an off-policyAdvantage-based Policy Transfer algorithm, APT-RL, for fixed domainenvironments. Its novelty is in using the popular notion of ``advantage'' as aregularizer, to weigh the knowledge that should be transferred from the source,relative to new knowledge learned in the target, removing the need forheuristic choices. Further, we propose a new transfer performance metric toevaluate the performance of our algorithm and unify existing transfer RLframeworks. Finally, we present a scalable, theoretically-backed tasksimilarity measurement algorithm to illustrate the alignments between ourproposed transferability metric and similarities between source and targetenvironments. Numerical experiments on three continuous control benchmark tasksdemonstrate that APT-RL outperforms existing transfer RL algorithms on mosttasks, and is $10\%$ to $75\%$ more sample efficient than learning fromscratch.

An advantage based policy transfer algorithm for reinforcement learning with metrics of transferability