Estimating Quantum Execution Requirements for FeatureSelection in Recommender Systems Using Extreme Value Theory

Jiayang Niu, Jie Li, KE DENG, Mark Sanderson, Yongli Ren

Abstract: Recent progress in quantum computing has advanced research in quantum assisted information retrieval and recommender systems, especially for feature selection via Quadratic Unconstrained Binary Optimization (QUBO). However, while existing work primarily focuses on effectiveness and efficiency, However, it often neglects the inherent noise and probabilistic nature of quantum hardware.In this paper, we propose a method based on Extreme Value Theory (EVT) to estimate the number of quantum executions (shots)needed to reliably obtain high-quality solutions—comparable to or better than classical baselines. Experiments on both simulators and two physical quantum devices demonstrate that our method effectively estimates the number of required runs to obtain satisfactory solutions on two widely used benchmark datasets.