Learning-based Optimal Quantum Switch Scheduling

In this paper, we consider the problem of optimal scheduling for quantum switches with dynamic demand and random entanglement successes. Different from prior results that often focus on (known) fixed entanglement success probabilities, we assume zero prior knowledge about the entanglement success probabilities and allow them to vary from time to time in an adversarial manner. We propose a learning-based algorithm QSSoftMW based on the framework developed in [1], which combines adversarial learning and Lyapunov queue analysis. We show that QSSoftMW is able to automatically adapt to the changing system statistics and ensure quantum switch stability.