Machine learning assisted construction of a shallow depth dynamic ansatz for noisy quantum hardware

Area of Science:

• Quantum Computing
• Computational Chemistry
• Machine Learning

Background:

• Noisy Intermediate-Scale Quantum (NISQ) hardware enables molecular simulations via dynamic ansatz construction.
• Current ansatz construction methods require high measurement costs, limiting practical applications.
• Developing resource-efficient quantum algorithms is crucial for near-term quantum computing.

Purpose of the Study:

• To develop a novel protocol for constructing expressive and shallow quantum simulation ansatzes.
• To reduce measurement costs and enhance robustness against hardware noise in variational quantum eigensolver (VQE) algorithms.
• To facilitate accurate molecular property determination on NISQ devices.

Main Methods:

• Utilized regenerative machine learning and many-body perturbation theory to identify dominant excited determinants.
• Trained machine learning models on low-rank expansions of the N-electron Hilbert space.
• Incorporated selected excited determinants into the ansatz via low-rank decomposition.

Main Results:

• Achieved a significant reduction in quantum measurement costs and ansatz depth.
• Demonstrated robustness towards hardware noise through numerical simulations.
• The proposed method is compatible with neural error mitigation techniques.

Conclusions:

• The developed resource-efficient approach is essential for molecular simulations on NISQ hardware.
• Enables accurate determination of spectroscopic and molecular properties.
• Facilitates the study of novel chemical phenomena using near-term quantum computers.