Abstract:In this note, variational Monte Carlo method based on neural quantum states for spin systems is reviewed. Using a neural network as the wave function allows for a more generalized expression of various types of interactions, including highly non-local interactions, which are closely related to its non-linear activation functions. Additionally, neural networks can represent relatively complex wave functions with relatively small computational resources when dealing with higher-dimensional systems, which is undoubtedly a "flattening" advantage. In quantum-state tomography, the representation method of neural quantum states has already achieved significant results, hinting at its potential in handling larger-sized systems.
What problem does this paper attempt to address?
This paper mainly discusses the application of Variational Monte Carlo (VMC) method in quantum many-body physics and quantum chemistry, especially how to optimize this method using Neural Quantum States. VMC is based on the variational principle, estimating the ground-state energy of a quantum system by optimizing a parameterized wave function. The paper reviews the development of VMC, including McMillan's early work, Anderson's application in quantum chemistry, and Sorella's proposed Stochastic Reconfiguration (SR) method.
The paper proposes the use of deep learning techniques, especially neural networks, in VMC method to solve quantum many-body problems. Neural Quantum States are wave functions represented by neural networks, which can more efficiently express different types of interactions, especially in handling high-dimensional systems, complex wave functions can be represented with relatively small computational resources. The authors also discuss importance sampling methods, such as using Markov chains for random sampling, and introduce the Metropolis algorithm combined with local updates to generate configurations closer to the wave function.
In addition, the paper introduces the Stochastic Reconfiguration technique for optimizing parameters, which is a method to find the variational wave function closest to the ground state. By establishing a linear equation system related to time evolution, the parameters can be iteratively updated to gradually approach the system's ground state.
Finally, the paper briefly introduces Neural Quantum States (NQS), a representation form of wave functions suitable for the variational part of VMC method. NQS utilizes the non-linear capability of neural networks to approximate multi-body quantum states, providing a new approach to solve quantum many-body problems, especially in scenarios where classical methods encounter difficulties, such as the sign problem in Quantum Monte Carlo or the limitations of compressed representation methods.
Overall, this paper aims to improve the Variational Monte Carlo method by combining modern machine learning techniques, especially neural networks, to more efficiently simulate and understand the behavior of complex quantum systems.