JSAI2022

Presentation information

General Session

General Session » GS-2 Machine learning

[2C4-GS-2] Machine learning: reinforcement learning (1)

Wed. Jun 15, 2022 1:20 PM - 3:00 PM Room C (Room C-2)

座長:谷本 啓(NEC)[現地]

2:40 PM - 3:00 PM

[2C4-GS-2-05] Generalization of Some Reinforcement Learning Formulae to Maximum Entropy Reinforcement Learning

〇Yuki Nakaguchi1 (1. NEC)

Keywords:Reinforcement Learning, Maximum Entropy

Recently, reinforcement learning (RL) has shown increasingly high performance in a variety of complex tasks of decision making and control. Especially, various merits and properties of its generalized formulation, Maximum Entropy RL (MaxEnt RL), have been revealed. Due to the introduction of the entropy regularization term, however, existing formulae for RL cannot be applied to MaxEnt RL without modification in general, which prevents the development of new algorithms and theoretical analysis of MaxEnt RL. To overcome this problem, we generalize some existing RL formulae to MaxEnt RL, while we give an organized review of MaxEnt RL.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password