Bonn 2025 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

QI: Fachverband Quanteninformation

QI 10: Quantum Machine Learning II

QI 10.7: Talk

Tuesday, March 11, 2025, 12:30–12:45, HS VIII

An SPSA-based Adaptive Shot Optimizer for variational algorithms — •Matteo Antonio Inajetovic and Anna Pappa — Technische Universität Berlin, Berlin, Germany

Adaptive shot optimizers dynamically adjust shot budget based on gradient variance, ensuring efficient shot allocation and significantly reducing the number of shots required for variational quantum algorithms. This is especially critical for concrete applications on noisy intermediate-scale quantum (NISQ) devices, where limited hardware access and high measurement costs pose substantial challenges. This work introduces adaptiveSPSA, a novel optimization method combining Simultaneous Perturbation Stochastic Approximation (SPSA) with adaptive shot strategies. Unlike other shot-frugal optimizers that rely on parameter-shift rules, adaptiveSPSA, leveraging the inherent efficiency of SPSA, computes gradient estimates using only two circuit executions per optimization step. Therefore, the proposed work is more robust to problem scaling, as the parameter-shift rule requires a number of gradient evaluations that scales linearly with the number of parameters, whereas SPSA maintains a constant number of evaluations regardless of parameter count. Numerical experiments on the Quantum Approximate Optimization Algorithm benchmark demonstrate that adaptiveSPSA outperforms Rosalin, one of the state-of-the-art methods, achieving superior performance still using a small amount of shots. These results underscore its potential to enhance the scalability and efficiency of variational quantum algorithms in practical applications with nowadays devices.

Keywords: Quantum Computing; Variational Quantum Algorithms; Optimization