Bounded Rationality Equilibrium Learning in Mean Field Games

Authors: Yannick Eich, Christian Fabian, Kai Cui, Heinz Koeppl

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section we evaluate our algorithms and analyze the different equilibria for several MFGs. We analyze the efficacy of our methods for a Susceptible-Infectious Susceptible (SIS) problem and a sequential version of Rock-Paper-Scissor (RPS) game. Additionally, we evaluate random MFGs, similar to P erolat et al. (2022), by creating random transition and reward matrices and adding a mean-field dependent function to the reward that promotes swarm avoiding behaviour. Detailed game descriptions are found in Eich et al. (2024, Appendix G). For code, see https://github.com/yannickeich/QRE-MFG. To measure algorithm efficiency, we quantify the distance...
Researcher Affiliation Academia Yannick Eich, Christian Fabian, Kai Cui, Heinz Koeppl Dept. of Electrical Engineering and Information Technology, Technische Universit at Darmstadt EMAIL
Pseudocode Yes Algorithm 1: Generalized Fixed-Point Iteration (GFPI). Algorithm 2: Generalized Fictitious Play (GFP). Algorithm 3: Sequential RH-GFP.
Open Source Code Yes For code, see https://github.com/yannickeich/QRE-MFG.
Open Datasets No In this section we evaluate our algorithms and analyze the different equilibria for several MFGs. We analyze the efficacy of our methods for a Susceptible-Infectious Susceptible (SIS) problem and a sequential version of Rock-Paper-Scissor (RPS) game. Additionally, we evaluate random MFGs, similar to P erolat et al. (2022), by creating random transition and reward matrices and adding a mean-field dependent function to the reward that promotes swarm avoiding behaviour. Detailed game descriptions are found in Eich et al. (2024, Appendix G).
Dataset Splits No The paper does not describe typical machine learning experiments with train/test/validation splits. The experiments focus on evaluating algorithms in simulated game environments, not on partitioning an existing dataset.
Hardware Specification No The paper does not mention any specific hardware specifications used for running the experiments.
Software Dependencies No The paper does not provide specific software names with version numbers.
Experiment Setup Yes Figure 2: Convergence of GFP for the Susceptible-Infectious Susceptible MFG with α = 1.0 and β = 0.95.