Learning against learning : evolutionary dynamics of reinforcement learning algorithms in strategic interactions

Abstract

People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the "Taverne" license above, please follow below link for the End User Agreement:

Task-Oriented Satellite-UAV Networks With Mobile-Edge Computing

“…It is concluded from ( 44) and ( 45) that, when Similarly, for Agent u,v,2 , based on [44], [46], we can derive…”

Analysis of Evolutionary Dynamics for Bidding Strategy Driven by Multi-Agent Reinforcement Learning

“…However, the factors affecting the converged bidding price are implicitly indicated therein. In [5], the numerical connection between EGT with RDEs and some baseline MARL algorithms is proved, implying that EGT with RDEs can explicitly reveal the factors affecting the converged result in MARL. Thus, in this letter, the correlation between WoLF-PHC and EGT is investigated and adopted to analyse the learning dynamics.…”

“…For EPs, the state refers to [𝜆 , , , 𝑃 , , ], the action is [ 𝑃 , , 𝜆 , ]. The EGT with RDEs, instead, presents the change of probability of multiple "players" selecting different "strategies", and these players will imitate the strategy of those who obtain the largest "payoff" [5]. Empirically, the strategy can be considered as the principle of selecting actions.…”

“…The numerical connection between the EGT and MARL lies in the speed of change of strategies and actions. If the speed of "strategy change" can be proved to be proportional to that of "policy change", then the expression of policy change can be directly replaced by that of strategy change [5], in which the factors affecting the converged result are explicitly indicated.…”

“…For EPs taking the price-maker strategy, they tend to submit a higher bidding price, to pursue more benefit while taking the risk of failure of bidding. Based on the derivations presented in [5], the change of probability of player 𝑥 and 𝑦 [10] selecting different strategies 𝑝 and 𝑞 can be written as:…”