Zero-sum Polymatrix Markov Games: Equilibrium Collapse and Efficient Computation of Nash Equilibria

Authors: Fivos Kalogiannis, Ioannis Panageas

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical The proof relies on the fact that zero-sum polymatrix Markov games with a switching controller have the following important property: the marginals of a coarse-correlated equilibrium constitute a Nash equilibrium (see Section 3.2). We refer to this phenomenon as equilibrium collapse. This property was already known for zero-sum polymatrix normal-form games by Cai et al. (2016) and our results generalize the aforementioned work for Markov games. As a corollary, we get that any algorithm in the literature that guarantees convergence to approximate coarse-correlated equilibria Markovian policies e.g., (Daskalakis et al., 2022) can be used to get approximate Nash equilibria.
Researcher Affiliation Academia Fivos Kalogiannis Department of Computer Science University of California, Irvine Irvine, CA fkalogia@uci.edu Ioannis Panageas Department of Computer Science University of California, Irvine Irvine, CA ipanagea@ics.uci.edu
Pseudocode Yes Algorithm 1 SPo CMAR (Daskalakis et al., 2022) and Algorithm 2 Est Visitation are provided in Appendix D.
Open Source Code No The paper refers to algorithms from other papers but does not state that the authors are providing open-source code for their own work or methodology.
Open Datasets No The paper is theoretical and does not involve training models on datasets.
Dataset Splits No The paper is theoretical and does not involve dataset splits for validation.
Hardware Specification No The paper is theoretical and does not describe experiments, thus no hardware specifications are mentioned.
Software Dependencies No The paper is theoretical and focuses on mathematical proofs and algorithms. It does not mention specific software dependencies or versions for implementation.
Experiment Setup No The paper is theoretical and does not describe empirical experiments, therefore, no experimental setup details like hyperparameters or training settings are provided.