This open-source implementation considers a machine learning (ML) algorithm for predicting the output properties of an arbitrarily complex quantum process (the quantum process could even be an exponentially large quantum circuit!).
We require g++ (C++ compiler), python version 3, and Jupyter Notebook (https://ipython.org/notebook.html).
On the experimental side, we require preparation of product states and single-qubit Pauli measurements (i.e., each measurement measures all qubits on some Pauli X, Y, or Z- basis). This should be readily available in many quantum platforms.
An introduction to this ML algorithm and the underlying theory can be found in our papers: https://arxiv.org/abs/2210.14894
Every folder (except for Eigen/), such as 50spins-allZ-many-t-homogeneous or Sys-40spins-oneZ-allt-homogeneous, corresponds to a particular quantum system that we consider in the numerical experiments. The folder Eigen/ is an open-source library (https://eigen.tuxfamily.org/index.php?title=Main_Page) for performing eigendecomposition.
To create the executable files in each folder (XXZ or XXZ-more-general), type make in each folder (this requires the C++ compiler g++). Running the executable file ./XXZ creates states.txt and values.txt, which consist of the training data for the ML algorithm.
The training and prediction of the ML algorithm are given in LearningQuantumProcess.ipynb (this requires Jupyter Notebook). To open LearningQuantumProcess.ipynb, type ipython notebook in the main folder and click LearningQuantumProcess.ipynb on the webpage.