Estimating quantum discord for two-qubit quantum systems via machine learning

Abstract Quantum discord characterizes non-classical correlations between two or more subsystems that persist even in the absence of entanglement. Unlike quantum entanglement, this measure captures more general quantum correlations, yet its estimation remains challenging due to the inherent difficulty in determining the optimal measurement orientations. In this work, we present a machine learning approach to estimate quantum discord for two-qubit systems with the help of artificial neural networks. Two distinct input strategies are investigated: one involving complete quantum state information derived from the density matrix elements and the other involving partial information derived from the expected values of measurements. Numerical simulations demonstrate that both artificial neural network architectures can predict the quantum discord of new quantum states with high precision. Our findings provide a new way to estimate quantum correlations, demonstrating the effectiveness of machine learning in quantum information tasks.