研究目的
To validate large-scale quantum devices, specifically boson samplers, by distinguishing between those using indistinguishable photons and those that do not, using unsupervised machine-learning methods to identify pathologies.
研究成果
Pattern-recognition techniques, particularly K-means clustering, effectively distinguish between boson samplers using indistinguishable and distinguishable photons, outperforming previous methods. The approach is scalable to large systems and generalizable to other quantum sampling tasks, providing a valuable tool for validating quantum devices without requiring permanent calculations.
研究不足
The method does not provide a definitive test for boson sampler validity; it only rules out specific pathologies. It requires a trusted sample for comparison and may have reduced accuracy in non-ideal conditions, such as lossy scenarios or high-dimensional spaces where fine-tuning is challenging.
1：Experimental Design and Method Selection:
The study employs unsupervised machine-learning techniques, particularly K-means clustering, to analyze output distributions from boson samplers. The protocol involves comparing samples from a trusted boson sampler and an untrusted device using clustering and a chi-square test.
2：Sample Selection and Data Sources:
Data are generated from numerical simulations of boson samplers with n photons in m modes (e.g., n=3, m=13) and experimental setups using integrated photonic chips. Samples include outputs from indistinguishable and distinguishable photons.
3：List of Experimental Equipment and Materials:
Integrated photonic chips fabricated via femtosecond laser writing, delay lines for adjusting photon distinguishability, parametric down-conversion sources, polarizing beam splitters, polarization controllers, fiber arrays, interferential filters, and beta barium borate crystals.
4：Experimental Procedures and Operational Workflow:
For numerical simulations, samples are generated using algorithms for indistinguishable and distinguishable photons. For experiments, photons are injected into photonic chips, and output states are collected. The clustering algorithm (e.g., K-means) is applied to partition data, followed by a chi-square test to compare sample compatibility.
5：Data Analysis Methods:
Statistical analysis includes chi-square tests for compatibility, with success percentages calculated from multiple trials. Correlation coefficients (Pearson and Spearman) are used to analyze distribution structures.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
