Quantum Similarity Testing with Convolutional Neural Networks.

Area of Science:

• Quantum Information Science
• Machine Learning
• Quantum Computing

Background:

• Benchmarking quantum devices is crucial for advancing quantum computing and simulation.
• Comparing continuous variable quantum states, especially non-Gaussian ones, presents a significant challenge.
• Existing methods are insufficient for similarity testing of uncharacterized continuous variable quantum states.

Purpose of the Study:

• To develop a machine learning algorithm for comparing unknown continuous variable quantum states.
• To enable similarity testing for non-Gaussian quantum states using limited and noisy data.
• To provide a method for benchmarking near-term quantum computers and simulators.

Main Methods:

• Utilized a convolutional neural network (CNN) for quantum state similarity assessment.
• Developed a lower-dimensional state representation from measurement data.
• Trained the CNN using simulated, experimental, or combined data from fiducial states.

Main Results:

• Successfully tested the algorithm on noisy cat states and states from arbitrary selective number-dependent phase gates.
• Demonstrated the algorithm's ability to compare continuous variable states across different experimental platforms.
• Showcased the network's applicability to testing state equivalence up to Gaussian unitary transformations.

Conclusions:

• The developed machine learning algorithm effectively compares unknown continuous variable quantum states.
• This approach overcomes limitations in testing non-Gaussian states and noisy data.
• The method offers a powerful tool for benchmarking quantum devices and advancing quantum information science.