Payoff-based learning with matrix multiplicative weights in quantum games
This work addresses the challenge of limited information in quantum game theory, offering incremental algorithmic improvements for researchers in quantum computing and optimization.
The paper tackles the problem of learning in quantum games with minimal scalar feedback, introducing a suite of minimal-information matrix multiplicative weights (3MW) methods that adapt to different information frameworks. The results show that the 3MW method achieves an O(1/√T) convergence rate in quantum min-max games with deterministic feedback and an O(T^{-1/4}) rate with random feedback, while a regularized variant ensures local convergence to stable equilibria in non-zero-sum games.
In this paper, we study the problem of learning in quantum games - and other classes of semidefinite games - with scalar, payoff-based feedback. For concreteness, we focus on the widely used matrix multiplicative weights (MMW) algorithm and, instead of requiring players to have full knowledge of the game (and/or each other's chosen states), we introduce a suite of minimal-information matrix multiplicative weights (3MW) methods tailored to different information frameworks. The main difficulty to attaining convergence in this setting is that, in contrast to classical finite games, quantum games have an infinite continuum of pure states (the quantum equivalent of pure strategies), so standard importance-weighting techniques for estimating payoff vectors cannot be employed. Instead, we borrow ideas from bandit convex optimization and we design a zeroth-order gradient sampler adapted to the semidefinite geometry of the problem at hand. As a first result, we show that the 3MW method with deterministic payoff feedback retains the $\mathcal{O}(1/\sqrt{T})$ convergence rate of the vanilla, full information MMW algorithm in quantum min-max games, even though the players only observe a single scalar. Subsequently, we relax the algorithm's information requirements even further and we provide a 3MW method that only requires players to observe a random realization of their payoff observable, and converges to equilibrium at an $\mathcal{O}(T^{-1/4})$ rate. Finally, going beyond zero-sum games, we show that a regularized variant of the proposed 3MW method guarantees local convergence with high probability to all equilibria that satisfy a certain first-order stability condition.