This paper introduces a quantum state tomography scheme using convolutional neural networks (CNNs) to approximate the probability distribution of informationally complete measurement outcomes. The method achieves high classical fidelities, outperforming standard maximum likelihood estimation (MLE), and reduces observable estimation errors by up to an order of magnitude. The number of variational parameters scales polynomially with system size, enabling efficient reconstruction of states.

Tobias Schmale, Moritz Reh, Martin Gärttner