Statistical Analysis of Reinforcement Learning Training

Maximilian Moll (), Matthias Schilling () and Stefan Pickl ()
Additional contact information
Maximilian Moll: Institute of Operations Research, University of the Bundeswehr Munich
Matthias Schilling: Institute of Operations Research, University of the Bundeswehr Munich
Stefan Pickl: Institute of Operations Research, University of the Bundeswehr Munich

Chapter Chapter 57 in Operations Research Proceedings 2023, 2025, pp 447-452 from Springer

Abstract: Abstract One of the most urgent challenges in Reinforcement Learning research is the lack of reproducibility. Therefore, to further the understanding of the training behavior of Reinforcement Learning agents, we analyze the training of agents playing the established baseline environment Taxi. In particular, we contrast results based on different forms of exploration. In addition, we can demonstrate that in this context penalization without termination is to be the preferred punishment for incorrect actions.

Keywords: Reinforcement learning; Q learning; Statistical testing (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:lnopch:978-3-031-58405-3_57

Ordering information: This item can be ordered from
http://www.springer.com/9783031584053

DOI: 10.1007/978-3-031-58405-3_57

Access Statistics for this chapter

More chapters in Lecture Notes in Operations Research from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().