Lévy noise promotes cooperation in the prisoner’s dilemma game with reinforcement learning

Uncertainties are ubiquitous in everyday life, and it is thus important to explore their effects on the evolution of cooperation. In this paper, the prisoner’s dilemma game with reinforcement learning subject to Lévy noise is studied. Specifically, diverse fluctuations mimicked by Lévy distributed n...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Nonlinear dynamics Ročník 108; číslo 2; s. 1837 - 1845
Hlavní autori: Wang, Lu, Jia, Danyang, Zhang, Long, Zhu, Peican, Perc, Matjaž, Shi, Lei, Wang, Zhen
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Dordrecht Springer Netherlands 01.04.2022
Springer Nature B.V
Predmet:
Algorithms
Automotive Engineering
Classical Mechanics
Control
Cooperation
Dynamical Systems
Engineering
Evolution
Game theory
Machine learning
Mechanical Engineering
Noise
Original Paper
Random noise
Strategy
Vibration
learning
Lévy noise
Prisoner’s dilemma
Self-regarding
Evolutionary dynamics
Cooperation
ISSN:0924-090X, 1573-269X
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Uncertainties are ubiquitous in everyday life, and it is thus important to explore their effects on the evolution of cooperation. In this paper, the prisoner’s dilemma game with reinforcement learning subject to Lévy noise is studied. Specifically, diverse fluctuations mimicked by Lévy distributed noise are reflected in the payoff matrix of each player. At the same time, the self-regarding Q -learning algorithm is considered as the strategy update rule to learn the behavior that achieves the highest payoff. The results show that not only does Lévy noise promote the evolution of cooperation with reinforcement learning, it does so comparatively better than Gaussian noise. We explain this with the iterative updating pattern of the self-regarding Q -learning algorithm, which has an accumulative effect on the noise entering the payoff matrix. It turns out that under Lévy noise, the Q -value of cooperative behavior becomes significantly larger than that of defective behavior when the current strategy is defection, which ultimately leads to the prevalence of cooperation, while this is absent with Gaussian noise or without noise. This research thus unveils a particular positive role of Lévy noise in the evolutionary dynamics of social dilemmas.
Bibliografia:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0924-090X
1573-269X
DOI:10.1007/s11071-022-07289-7