Towards Redefining the Reproducibility in Quantum Computing: A Data Analysis Approach on NISQ Devices
- Resource Type
- Conference
- Authors
- Senapati, Priyabrata; Wang, Zhepeng; Jiang, Weiwen; Humble, Travis S; Fang, Bo; Xu, Shuai; Guan, Qiang
- Source
- 2023 IEEE International Conference on Quantum Computing and Engineering (QCE) QCE Quantum Computing and Engineering (QCE), 2023 IEEE International Conference on. 01:468-474 Sep, 2023
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
Engineering Profession
Computers
Performance evaluation
Analytical models
Quantum computing
Quantum algorithm
Computational modeling
Machine learning
quantum computing
quantum machine learning
reproducibility
quantum noise
- Language
Although the building of quantum computers has kept making rapid progress in recent years, noise is still the main challenge for any application to leverage the power of quantum computing. Existing works addressing noise in quantum devices proposed noise reduction when deploying a quantum algorithm to a specified quantum computer. The reproducibility issue of quantum algorithms has been raised since the noise levels vary on different quantum computers. Importantly, existing works largely ignore the fact that the noise of quantum devices varies as time goes by. Therefore, reproducing the results on the same hardware will even become a problem. We analyze the reproducibility of quantum machine learning (QML) algorithms based on daily model training and execution data collection. Our analysis shows a correlation between our QML models' test accuracy and quantum computer hardware's calibration features. We also demonstrate that noisy simulators for quantum computers are not a reliable tool for quantum machine learning applications.