Transformer neural networks and quantum simulators: a hybrid approach for simulating strongly correlated systems

Hannah Lange,Guillaume Bornet,Gabriel Emperauger,Cheng Chen,Thierry Lahaye,Stefan Kienle,Antoine Browaeys,Annabelle Bohrdt

2024-06-01

Abstract:Owing to their great expressivity and versatility, neural networks have gained attention for simulating large two-dimensional quantum many-body systems. However, their expressivity comes with the cost of a challenging optimization due to the in general rugged and complicated loss landscape. Here, we present a hybrid optimization scheme for neural quantum states (NQS) that involves a data-driven pretraining with numerical or experimental data and a second, Hamiltonian-driven optimization stage. By using both projective measurements from the computational basis as well as expectation values from other measurement configurations such as spin-spin correlations, our pretraining gives access to the sign structure of the state, yielding improved and faster convergence that is robust w.r.t. experimental imperfections and limited datasets. We apply the hybrid scheme to the ground state search for the 2D transverse field Ising model and the 2D dipolar XY model on $6\times 6$ and $10\times 10$ square lattices with a patched transformer wave function, using numerical and experimental data from a programmable Rydberg quantum simulator [Chen et al., Nature 616 (2023)], with snapshots of the quantum system obtained from the different measurement configurations, and show that the information from the second basis highly improves the performance. Our work paves the way for a reliable and efficient optimization of neural quantum states.

Disordered Systems and Neural Networks,Quantum Gases,Strongly Correlated Electrons,Quantum Physics

What problem does this paper attempt to address?

The paper primarily aims to address a significant challenge in quantum many-body physics: simulating large, strongly correlated 2D quantum systems. The precise parameterization of such systems becomes infeasible due to the exponential growth of the Hilbert space with system size. To tackle this issue, the authors propose a hybrid optimization scheme to train Neural Quantum States (NQS), specifically using the Transformer neural network architecture.

Specifically, the scheme comprises two stages:
1. **Data-driven pretraining**: Utilizing data from experiments or numerical simulations for pretraining. This data can be projection measurement results obtained from the computational basis (Z-basis) or expectation values from other measurement configurations (such as spin-spin correlations). In this way, the symbolic structure information of the quantum state under study can be obtained, thereby accelerating the convergence process and making it more robust to experimental imperfections and limited datasets.
2. **Hamiltonian-driven optimization**: After pretraining, the NQS is further optimized using the variational Monte Carlo method to minimize the system's energy.

The hybrid optimization scheme mentioned in the paper is applied to two specific models: the 2D Transverse Field Ising Model (TFIM) and the 2D Dipolar XY Model. Compared to training based solely on the variational Monte Carlo method, this approach significantly improves the quality of the optimization results, especially when combining different measurement configurations (Z-basis and X-basis), greatly enhancing the optimization speed. Additionally, the authors found that even with very limited data, this hybrid scheme can achieve good results.

Overall, the goal of the paper is to propose an efficient and reliable NQS optimization method to facilitate the simulation of more complex quantum systems, such as fermionic quantum states or quantum spin liquids.