arXiv:2609.21944
Guiding Agents of Quantum Games to Equilibrium using Matrix Exponential Fixed-Point Iteration
Alireza Habibi, Luis F. Abanto Leon, Setareh Maghsudi
Abstract
In recent years, quantum game theory has gained significant attention as a framework for studying decision-making in multi-agent systems using quantum principles. However, computing equilibrium strategies is challenging because the dimension of the joint Hilbert space grows as the product of the players' local dimensions. In this paper, we consider an extended Gutoski-Watrous (EGW) game in which each player's quantum strategy is represented by a local density matrix. We derive tensor-contraction expressions for the payoff functions and their gradients, thereby avoiding the explicit construction of the full joint density matrix and its computationally expensive multiplication by the payoff operators. Building on the resulting effective Hamiltonians, we propose the Matrix Exponential Fixed-Point Iteration with Annealing (MEFPIA) algorithm to search for equilibrium points in EGW games. We compare MEFPIA with the Matrix Multiplicative Weights Update (MMWU) algorithm in terms of convergence. For the tested instances and parameter settings, both algorithms approach the same strategy profiles and payoffs, while MEFPIA achieves lower relative error in fewer iterations. These results indicate that MEFPIA is a promising numerical method for equilibrium search in multi-agent quantum games. Our findings provide important insights into the quantum game theory's potential for addressing complex decision-making processes, as well as opening up new paths for future research and exploration in multi-agent quantum systems.
AI Summary
The paper studies numerical Nash-equilibrium search in multiplayer noninteractive quantum games, where each player chooses a local density matrix and payoffs are expectations of joint reward operators on the product strategy profile. It derives tensor-contraction expressions for payoffs and effective local Hamiltonians, avoiding explicit construction of the joint density matrix. Relative to evaluating full matrix products, the stated payoff-evaluation cost decreases from $O(\prod_i d_i^3)$ to $O((N+1)\prod_i d_i^2)$; sparse reward operators also permit reduced storage. The proposed Matrix Exponential Fixed-Point Iteration with Annealing (MEFPIA) simultaneously updates players using entropy-regularized best responses, $\rho_i\propto\exp(-H_i/T)$, while decreasing $T$. The Gibbs response is established as the unique regularized optimum using quantum relative entropy. If the strategy iterates converge while $T\to0$, their limit is a Nash equilibrium; convergence of the iteration itself is not established. Numerical comparisons with matrix multiplicative weights use 100 random initial profiles. Tests include two-qubit Prisoner’s-Dilemma variants with product or entangled projective measurements, and a three-player game with local dimension five and 27 product POVM outcomes. In the tested settings, MEFPIA reaches lower relative payoff error in fewer iterations. The simplest test reaches approximately $10^{-16}$ error within 100 iterations; the entangled-measurement test reports payoff $2.25$ per player. The larger test uses commuting local POVMs and yields mixed limiting states with purity approximately $1/2$. The evidence is restricted to these instances and parameter choices. Cooling-rate sensitivity is examined, while convergence properties, larger games, and more general measurement structures remain future work.
ChatGPT and QuillBot were used to improve the language and readability of this paper.
Central claim may not hold: In Experiment 1's Bell-outcome game, the reported pure limiting strategies with payoff 2.25 cannot be a Nash equilibrium because each player can obtain at least 2.5 by best responding to the other's pure strategy. The relative-payoff-error metric cannot certify equilibrium or convergence to the same strategy profiles, and the specified Bell-outcome game admits profiles with zero reported payoff error but a profitable unilateral deviation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.