SantaQlaus: A resource-efficient method to leverage quantum noise for optimization of variational quantum algorithms

We introduce SantaQlaus, a resource-efficient optimizer for variational quantum algorithms (VQAs), including VQE and QML applications. Classical optimization in VQAs often faces complex landscapes of local minima and saddle points. While some quantum-aware optimizers adaptively adjust shot numbers, their focus is mainly on gain per iteration, not on leveraging quantum noise strategically. SantaQlaus, inspired by the classical Santa algorithm, explicitly uses quantum noise for optimization. The algorithm dynamically adjusts shot numbers within an annealing framework: fewer shots are used early for efficient resource use and landscape exploration, while more are used later for greater accuracy. Numerical simulations on VQE and QML show that SantaQlaus outperforms existing methods, mitigating risks of poor local optima while maintaining shot efficiency. This paves the way for efficient and robust training of quantum variational models.