In the first part of my talk I will give an introduction to the concept of neural-network quantum states (NQSs) used as variational functions in the Monte Carlo studies of frustrated quantum spin systems. I will focus on basic principles of the method, the general motivation to use these non-standard variational functions, as well as about their relation with machine learning techniques. 

In the second part of the talk I will discuss in detail our recent results sumarized in Ref.[1] where we have utilized neural-network quantum states (NQSs) to investigate the ground-state properties of the Heisenberg model on a Shastry-Sutherland lattice via variational Monte Carlo method. We have shown that already relatively simple NQSs can be used to approximate the ground state of this model in its different phases and regimes. We have compared several types of NQSs with each other on small lattices and benchmarked their variational energies against the exact diagonalization results. We have then argued that when precision, generality and computational costs are taken into account, a good choice for addressing larger systems is a shallow restricted Boltzmann machine NQS. I will show during the talk that such NQS can describe the main phases of the model in zero magnetic field. Moreover, NQS based on a restricted Boltzmann machine correctly describes the intriguing plateaus forming in magnetization of the model as a function of increasing magnetic field.

[1] M. Mezera, J. Menšíková, P. Baláž, M. Žonda, Neural Network Quantum States analysis of the Shastry-Sutherland model arXiv:2303.14108 (2023)