Quantum generative adversarial learning in photonics.

Area of Science:

• Quantum computing
• Machine learning
• Quantum photonics

Background:

• Quantum generative adversarial networks (QGANs) merge quantum computing and machine learning.
• Investigating QGAN performance on noisy intermediate-scale quantum (NISQ) devices is crucial.
• Understanding the impact of noise and defects on QGANs is essential for practical applications.

Purpose of the Study:

• To experimentally demonstrate a QGAN model in photonics for the first time.
• To investigate the effects of noise and defects on QGAN performance.
• To assess the feasibility of QGANs on NISQ-era quantum hardware.

Main Methods:

• Utilized a programmable silicon quantum photonic chip.
• Implemented and tested a QGAN model on the photonic chip.
• Introduced controlled noise and simulated defects in the QGAN components.

Main Results:

• Achieved high-fidelity quantum data generation (>90%).
• Demonstrated robust QGAN performance despite significant damage to phase shifters (up to 50%).
• Showed effective operation even with substantial phase noise (up to 0.04π).

Conclusions:

• QGANs can successfully perform learning tasks on near-term quantum devices.
• The developed QGAN model exhibits resilience to noise and defects.
• This work validates the potential of QGANs for practical implementation on NISQ hardware.