Speaker 

Title

Generalizing Tensor Network Methods to Neural Network Wave Functions using Variational Monte Carlo

Abstract

Variational Monte Carlo (VMC) have recently regained considerable interest due to the advent of machine learning in combination with neural networks as variational states. Similar to the tensor network ansatz, neural networks are agnostic to the physics and can faithfully represent quantum many-body states. However, their application has been mostly restricted to ground state calculations. In this talk, I will present my work on developing Monte Carlo based variational techniques to simulate finite temperature properties and spectral functions using ideas originally derived in the context of tensor network methods. We implemented minimally entangled typical thermal states(METTS) to calculate thermal expectation values and generalized both the vector correction and Chebyshev expansion methods to calculate spectral functions.  We demonstrate these ideas for the 1D and 2D Heisenberg model using restricted Boltzmann machines (RBM) as variational wave functions.

Join Zoom Meeting

https://oist.zoom.us/j/94461527312
Meeting ID: 944 6152 7312