Abstract
We investigate machine learning-based noise classification aimed at the recognition of the Markovian character of a dynamics and the identification of correlations of classical noise, as well as their interplay in small quantum networks. We operate control based on Coherent Tunneling by Adiabatic Passage (CTAP) or Stimulated Raman Adiabatic Passage (STIRAP) in a three-level system using different pulse configurations as inputs to train a feedforward neural network. Our results show that supervised learning can classify distinct types of classical diagonal noise affecting the system. Three non-Markovian (quasistatic correlated, anti-correlated, and uncorrelated) and Markovian noise mechanisms are classified with $99%$ accuracy. Instead, correlations of Markovian noises cannot be classified with our method. The approach is robust against statistical measurement errors keeping its effectiveness even for physical measurements where a limited number of samples is available.
