MACHINE LEARNING PHASES OF MATTER IN OPEN QUANTUM SYSTEMS

Abstract

One of the main challenges in strongly correlated systems is the study of dissipative phases of matter. This thesis studies an emerging approach to this problem, which is the application of artificial neural networks (ANNs) combined with machine learning techniques. The ANN can be trained to learn the steady state of the system via a variational Monte Carlo scheme. We also propose a new ANN structure that respects the symmetry of the physical system and implement it in the case of translation invariance. We apply our method to the 1D and 2D dissipative transverse field Ising model, the 1D driven-dissipative XY model and the 1D driven-dissipative Bose-Hubbard model, obtaining overall satisfying results, which are also novel by the completion of this work, to the best of our knowledge. This study successfully extends the ANN approach to new physical models, showing great potential in this field of research.