Neural network based Variational Monte Carlo studies of the Hubbard Model

The Fermi Hubbard Model plays an essential role in our understanding of strongly correlated fermionic systems. However, conventional numerical techniques exhibit hard-to-overcome issues such as exponential growth of the Hilbert space or the fermion 'sign problem’. Previous studies have shown that recurrent neural networks can serve as reliable ansatz for the ground state wave functions of spin systems or the t-J models, overcoming some of the issues related to scaling. In this talk, we discuss our current efforts to exploit recurrent neural networks in variational QMC to represent the ground state of the square lattice Fermi Hubbard. We present our results on the use of quantum gas microscopy projective measurements to enhance convergence, and discuss future directions.