Geometry of policy improvement

DOI10.1007/978-3-319-68445-1_33MaRDI QIDQ1689145zbMATH OpenOpenAlexFDO

Authors Guido Montúfar, Johannes Rauh

Publication date 12 January 2018

Full work available at URL https://arxiv.org/abs/1704.01785

zbMATH Keywords

reinforcement learningpartially observable Markov decision processmemoryless stochastic policypolicy gradient theorem

Mathematics Subject Classification ID

Decision theory (91B06)

Abstract: We investigate the geometry of optimal memoryless time independent decision making in relation to the amount of information that the acting agent has about the state of the system. We show that the expected long term reward, discounted or per time step, is maximized by policies that randomize among at most $k$ actions whenever at most $k$ world states are consistent with the agent's observation. Moreover, we show that the expected reward per time step can be studied in terms of the expected discounted reward. Our main tool is a geometric version of the policy improvement lemma, which identifies a polyhedral cone of policy changes in which the state value function increases for all states.

Recommendations

• A polynomial time bound for Howard's policy improvement algorithm
• Finding optimal memoryless policies of POMDPs under the expected average reward criterion
• On the complexity of finite memory policies for Markov decision processes
• Near-optimal reinforcement learning in polynomial time
• Constrained Discounted Dynamic Programming

Cited in

(3)

• Algebraic optimization of sequential decision problems
• State observation accuracy and finite-memory policy performance
• scientific article; zbMATH DE number 5957207 (Why is no real title available?)

This page was built for publication: Geometry of policy improvement

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1689145)