Quantum State Tomography with Conditional Generative Adversarial Networks.

Area of Science:

• Quantum Information Science
• Machine Learning for Quantum Physics

Background:

• Quantum state tomography (QST) is essential for characterizing quantum devices but is computationally intensive.
• Existing QST methods struggle with the complexity of intermediate-scale quantum devices.

Purpose of the Study:

• To introduce a novel application of Conditional Generative Adversarial Networks (CGANs) for efficient Quantum State Tomography.
• To improve the fidelity and reduce the data and computational requirements of QST.

Main Methods:

• Augmented a CGAN with custom neural network layers to ensure physical density matrix output.
• Trained generator and discriminator networks iteratively using gradient-based methods.
• Demonstrated reconstruction of optical quantum states.

Main Results:

• QST-CGAN achieved high-fidelity reconstruction of optical quantum states.
• The method used orders of magnitude less data and iterative steps compared to projected-gradient and maximum-likelihood methods.
• Pretrained QST-CGAN can reconstruct quantum states in a single generator evaluation.

Conclusions:

• CGANs offer a powerful and efficient alternative for Quantum State Tomography.
• This approach significantly advances the characterization capabilities for intermediate-scale quantum systems.