Construction of Chemistry-Inspired Dynamic Ansatz Utilizing Generative Machine Learning.

Area of Science:

• Quantum Computing
• Computational Chemistry
• Machine Learning

Background:

• Generative machine learning models, such as Restricted Boltzmann Machines (RBMs), offer a viable strategy for constructing ansatz in quantum computing.
• Accurate determination of molecular energetics is crucial for understanding chemical phenomena.

Purpose of the Study:

• To develop an efficient method for constructing compact, chemistry-inspired ansatz using RBMs and many-body perturbation theory.
• To enable accurate calculation of molecular energetics for near-term quantum computers.

Main Methods:

• Leveraging RBMs trained on low-rank determinants to predict dominant high-rank determinants of the ground-state wave function.
• Constructing a shallow-depth ansatz incorporating these determinants after low-rank decomposition and perturbative screening.
• Utilizing Bayesian hyperparameter optimization for efficient RBM training with minimal data.

Main Results:

• The method successfully builds a compact ansatz by identifying and incorporating key determinants.
• No additional measurements are required beyond the initial training phase.
• Efficient performance is achieved with limited training data.

Conclusions:

• This RBM-based approach provides an efficient pathway to compute molecular properties accurately.
• The method facilitates the exploration of novel chemical phenomena using near-term quantum computing resources.