Neural network analytic continuation for Monte Carlo: Improvement by statistical errors

This study explores the use of neural network-based analytic continuation to extract spectra from Monte Carlo data. We apply this technique both synthetic and Carlo-generated The training sets for networks are carefully synthesized without “data leakage”. find that set should match input correlation functions in terms statistical error properties, such as noise level, dependence on imaginary time, time-displaced correlations. have developed a systematic method synthesize datasets. Our improved algorithm outperforms widely used maximum entropy highly noisy situations. As an example, our successfully extracted dynamic structure factor spin-1/2 Heisenberg chain quantum simulations.