Relatore: Stefano Martina (Università di Firenze)
Quantum Machine Learning (QML) is a new but very rapidly growing research and applied field where quantum information science meets machine learning algorithms. Here we are proposing to exploit artificial neural network models for quantum noise classification in stochastic quantum dynamics. In particular some Hamiltonian parameters (e.g. describing the evolution of a quantum simulator) are affected by noise, which is described as a stochastic process associated to a specific probability distribution. Our aim is to discriminate among different noise probability distributions and correlation parameters, by measuring the quantum state at discrete time instants. In particular, we choose classical machine learning algorithms as Support Vector Machines (SVM) and Recurrent Neural Networks (RNN), i.e. Long Short Term Memory (LSTM), and Gated Recurrent Unit (GRU), to perform supervised classification tasks on simulated quantum data. We believe that these results represent preliminary but very promising steps towards new QML algorithms for noise sensing and performance analysis in quantum computing, e.g. to detect Markovian vs. non-Markovian noise in real quantum systems (as optical and atomic platforms) currently exploited to build up the new powerful class of Noisy-Intermediate-Scale-Quantum (NISQ) HPC technologies.