Learning phases with Quantum Monte Carlo simulation cell

We propose a new machine learning input data type, "spin-opstring", derived from Stochastic Series Expansion Quantum Monte Carlo (QMC) simulations. It offers a compact, memory-efficient representation of QMC simulation cells, combining the initial state with an operator string that encodes the state's evolution through imaginary time. Using supervised machine learning on two models, we demonstrate the input's effectiveness in capturing both conventional and topological phase transitions. Additionally, we conduct a regression task to predict superfluid density, which reflects non-local properties of the quantum system, and achieve good accuracy. These results validate the spin-opstring as an effective input for machine learning applications.