همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Performance Analysis of Variational Quantum Classifiers for Classification Tasks in the NISQ Era

نویسندگان :

Saghar Kiani¹ Paria Kamalpour² Sarina Ghafouri³ Niloofar Mirzaei Chahardeh⁴

1- دانشگاه آزاد کرج 2- دانشگاه آزاد کرج 3- دانشگاه آزاد کرج 4- دانشگاه آزاد واحد علوم تحقیقات تهران

کلمات کلیدی :

quantum machine learning،variational quantum classifier،nisq،depolarizing noise،feature encoding

چکیده :

Recent advances in quantum computing have renewed interest in developing machine learning models capable of exploiting quantum phenomena such as superposition, entanglement, and exponentially large feature spaces. As current devices remain limited by noise and shallow circuit depths in the NISQ era, Variational Quantum Algorithms (VQAs) offer a practical path forward. This study provides a dual contribution: (1) a systematic review of Quantum Machine Learning (QML) research from 2020 to 2024, and (2) a controlled experimental evaluation of a two-qubit Variational Quantum Classifier (VQC) applied to the Iris dataset. A structured search across IEEE Xplore, ScienceDirect, and arXiv identified 90 papers, from which 30 met eligibility criteria. Experimentally, the VQC was implemented in Qiskit Aer using Angle Encoding, COBYLA optimization (maxiter=100, tol=1e-3), and a depolarizing noise channel with p=0.02. All experiments were repeated across 20 runs with a fixed seed (42). The VQC achieved an accuracy of 0.89 ± 0.01 and an F1-score of 0.88 ± 0.02, compared with 0.93 ± 0.02 and 0.91 ± 0.03 for an RBF-SVM baseline. Although numerical differences exist, a two-sample t-test confirmed that performance differences were not statistically significant (p = 0.04) at the 95% confidence level. Under noise, SVM accuracy dropped by ~7%, while the VQC degraded by only ~%3. These findings indicate that, while classical models still outperform quantum models on small datasets, the robustness and expressive capacity of VQAs highlight promising future potential for QML as hardware improvements are made.