Abstract:Strongly correlated quantum many-body states provide invaluable resources for state-of-the-art quantum information science, ranging from quantum simulation to quantum metrology. In this work, we propose a supervised machine learning algorithm for optimal control of quantum many-body atomic states in optical lattices by numerical simulations based on the Bayesian method of Gaussian process regression (GPR). We combine this method with the time evolving block decimation (TEBD) algorithm for the preparation of the Heisenberg antiferromagnetic state in a system of bosonic atoms confined with a one-dimensional optical lattice. The quantum many-body ground state of 80 atoms is efficiently optimized within a few hundred machine learning iterations, reaching a state fidelity above 96%. With a multistep learning strategy, we find the machine-learningbased optimal control method is scalable to large systems owing to its transferability. Its robustness against noise is demonstrated by considering imperfections that are typically present in optical lattice experiments. In the application of the GPR method to the preparation of the two-dimensional (2D) antiferromagnetic state, a state fidelity of 94% is reached for a 2D array of 6 Chi 6 spins through the time dependent variational principle (TDVP) algorithm, confirming the generalizability of our method. We further optimize the Hamiltonian ramping sequence crossing a quantum phase transition of the quantum XXZ spin chain with the exact diagonalization method, from which a control protocol for generating the long-sought atomic Greenberger-Horne-Zeilinger state is obtained. We believe the proposed optimal control method for quantum Hamiltonian ground-state preparation would benefit present ultracold-atom experiments in the study of strongly correlated many-body physics.

Bayesian learning for optimal control of quantum many-body states in optical lattices

Optimal control of linear Gaussian quantum systems via quantum learning control

Learning Spin Liquids on a Honeycomb Lattice with Artificial Neural Networks.

Learning in Quantum Control: High-Dimensional Global Optimization for Noisy Quantum Dynamics

Bayesian Optimization for Robust State Preparation in Quantum Many-Body Systems

High-dimensional reinforcement learning for optimization and control of ultracold quantum gases

Machine-learning-inspired quantum control in many-body dynamics

High-Dimensional Multi-fidelity Bayesian Optimization for Quantum Control

Bayesian optimization for state engineering of quantum gases

State Preparation in a Jaynes-Cummings Lattice with Quantum Optimal Control

Molecular Quantum Control Algorithm Design by Reinforcement Learning

Quantum optimal control of superconducting qubits based on machine-learning characterization

Large-scale quantum reservoir learning with an analog quantum computer

Machine-learning-inspired quantum optimal control of nonadiabatic geometric quantum computation via reverse engineering

Machine-Learning-Assisted Quantum Control in a Random Environment

Controlling nonergodicity in quantum many-body systems by reinforcement learning

The advantage of quantum control in many-body Hamiltonian learning

Reinforcement Learning for Many-Body Ground-State Preparation Inspired by Counterdiabatic Driving

Learning Robust and High-Precision Quantum Controls

Provably efficient machine learning for quantum many-body problems

Controlling Many-Body Quantum Chaos: Bose-Hubbard systems