Learning Explainable and Better Performing Representations of
POMDP Strategies

D 2024

Learning Explainable and Better Performing Representations of POMDP Strategies

BORK, Alexander; Debraj CHAKRABORTY; Kush GROVER; Jan KŘETÍNSKÝ; Stefanie MOHR et. al.

Základní údaje

Originální název

Learning Explainable and Better Performing Representations of POMDP Strategies

Autoři

BORK, Alexander; Debraj CHAKRABORTY; Kush GROVER; Jan KŘETÍNSKÝ a Stefanie MOHR

Vydání

Luxembourg City, Luxembourg, TACAS 2024, 30th International Conference on Tools and Algorithms for the Construction and Analysis of Systems, od s. 299-319, 21 s. 2024

Nakladatel

Springer

Další údaje

Jazyk

angličtina

Typ výsledku

Stať ve sborníku

Obor

10201 Computer sciences, information science, bioinformatics

Stát vydavatele

Švýcarsko

Utajení

není předmětem státního či obchodního tajemství

Forma vydání

tištěná verze "print"

Impakt faktor

Impact factor: 0.402 v roce 2005

Kód RIV

RIV/00216224:14330/24:00139094

Organizační jednotka

Fakulta informatiky

ISBN

978-3-031-57248-7

ISSN

DOI

https://doi.org/10.1007/978-3-031-57249-4_15

UT WoS

001284179800015

EID Scopus

2-s2.0-85192255092

Klíčová slova anglicky

POMDP; POMDP Strategies

Štítky

core_A, firank_A

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 31. 3. 2025 17:45, RNDr. Pavel Šmerk, Ph.D.

Anotace

V originále

Strategies for partially observable Markov decision processes (POMDP) typically require memory. One way to represent this memory is via automata. We present a method to learn an automaton representation of a strategy using a modification of the L∗-algorithm. Compared to the tabular representation of a strategy, the resulting automaton is dramatically smaller and thus also more explainable. Moreover, in the learning process, our heuristics may even improve the strategy’s performance. We compare our approach to an existing approach that synthesizes an automaton directly from the POMDP, thereby solving it. Our experiments show that our approach can lead to significant improvements in the size and quality of the resulting strategy representations.

Návaznosti

MUNI/I/1757/2021, interní kód MU

Název: MUNI Award in Science and Humanities (Akronym: Křetínský)

Investor: Masarykova univerzita, MUNI Award in Science and Humanities, MASH - MUNI Award in Science and Humanities

Citovat

BORK, Alexander; Debraj CHAKRABORTY; Kush GROVER; Jan KŘETÍNSKÝ a Stefanie MOHR. Learning Explainable and Better Performing Representations of POMDP Strategies. In TACAS 2024, 30th International Conference on Tools and Algorithms for the Construction and Analysis of Systems. Luxembourg City, Luxembourg: Springer, 2024, s. 299-319. ISBN 978-3-031-57248-7. Dostupné z: https://doi.org/10.1007/978-3-031-57249-4_15.

@inproceedings{2484767,
   author = {Bork, Alexander and Chakraborty, Debraj and Grover, Kush and Křetínský, Jan and Mohr, Stefanie},
   address = {Luxembourg City, Luxembourg},
   booktitle = {TACAS 2024, 30th International Conference on Tools and Algorithms for the Construction and Analysis of Systems},
   doi = {https://doi.org/10.1007/978-3-031-57249-4_15},
   keywords = {POMDP; POMDP Strategies},
   howpublished = {tištěná verze "print"},
   language = {eng},
   location = {Luxembourg City, Luxembourg},
   isbn = {978-3-031-57248-7},
   pages = {299-319},
   publisher = {Springer},
   title = {Learning Explainable and Better Performing Representations of POMDP Strategies},
   year = {2024}
}

TY  - CONF
ID  - 2484767
AU  - Bork, Alexander - Chakraborty, Debraj - Grover, Kush - Křetínský, Jan - Mohr, Stefanie
PY  - 2024
TI  - Learning Explainable and Better Performing Representations of POMDP Strategies
PB  - Springer
CY  - Luxembourg City, Luxembourg
SN  - 9783031572487
KW  - POMDP
KW  - POMDP Strategies
N2  - Strategies for partially observable Markov decision processes (POMDP) typically require memory. One way to represent this memory is via automata. We present a method to learn an automaton representation of a strategy using a modification of the L∗-algorithm. Compared to the tabular representation of a strategy, the resulting automaton is dramatically smaller and thus also more explainable. Moreover, in the learning process, our heuristics may even improve the strategy’s performance. We compare our approach to an existing approach that synthesizes an automaton directly from the POMDP, thereby solving it. Our experiments show that our approach can lead to significant improvements in the size and quality of the resulting strategy representations.
ER  -

BORK, Alexander; Debraj CHAKRABORTY; Kush GROVER; Jan KŘETÍNSKÝ a Stefanie MOHR. Learning Explainable and Better Performing Representations of POMDP Strategies. In \textit{TACAS 2024, 30th International Conference on Tools and Algorithms for the Construction and Analysis of Systems}. Luxembourg City, Luxembourg: Springer, 2024, s.~299-319. ISBN~978-3-031-57248-7. Dostupné z: https://doi.org/10.1007/978-3-031-57249-4\_{}15.

Přehled o publikaci