Policy Gradient Methods Find the Nash Equilibrium in N-player General-sum Linear-quadratic Games

Authors: Ben Hambly, Renyuan Xu, Huining Yang

JMLR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We illustrate our results with numerical experiments to show that even in situations where the policy gradient method may not converge in the deterministic setting, the addition of noise leads to convergence.
Researcher Affiliation Academia Mathematical Institute University of Oxford, Department of Industrial Systems and Engineering University of Southern California, Department of Operations Research and Financial Engineering Princeton University
Pseudocode Yes Algorithm 1 Natural Policy Gradient Method with Known Parameters
Open Source Code No The paper does not provide explicit access information or links to open-source code for the methodology described.
Open Datasets No We apply the natural policy gradient algorithm with unknown parameters to a two-player LQ game example with synthetic data consisting of a two-dimensional state variable and a one-dimensional control variable.
Dataset Splits No The paper uses synthetic data generated based on specified parameters and initial state distributions for its numerical experiments, rather than external datasets requiring explicit train/test/validation splits.
Hardware Specification No The paper does not provide specific hardware details (such as GPU or CPU models, or cloud resources) used for running its experiments.
Software Dependencies No The paper does not provide specific software dependency details (e.g., library or solver names with version numbers) used to replicate the experiments.
Experiment Setup Yes The natural policy gradient algorithm shows a reasonable level of accuracy within 1000 iterations (that is, the normalized error is less than 0.5%) for both players under different levels of system noise σ2, which ranges from 0 (deterministic dynamics) to 10. See Figure 1 for the case where r = 0.25 and Figure 2 for the case where r = 0.30.