Physics informed neural networks learning a two-qubit Hamiltonian

Machine learning techniques are employed to perform the full characterization of a quantum system. The particular artificial intelligence technique used to learn the Hamiltonian is called physics informed neural network (PINN). The idea behind PINN is the universal approximation theorem, which claims that any function can be approximate by a neural network if it contains enough complexity. Consequently, a neural network can be a solution of a physical model. Moreover, by means of extra data provided by the user, intrinsic physical parameters can be extracted from the approach called inverse-PINN. Here, we apply inverse-PINN with the goal of extracting all the physical parameters that constitutes a two qubit Hamiltonian. We find that this approach is very efficient. To probe the robustness of the inverse-PINN to learn the Hamiltonian of a two-qubit system, we use the IBM quantum computers as experimental platforms to obtain the data that is plugged in the PINN. We found that our method is able to predict the two-qubit parameters with 5% of accuracy on average.