Quantum extreme learning machines for photonic entanglement witnessing

Abstract: The estimation of properties of quantum states -- such as entanglement -- is a core need for the development of quantum technologies, yet remaining a demanding challenge. Standard approaches to property estimation rely on the modeling of the measurement apparatus and, often, a priori assumptions on their working principles. Even small deviations can greatly affect reconstruction accuracy and prediction reliability. Machine learning (ML) techniques have been proven promising to overcome these difficulties. However, interpretability issues related to overfitting limit the usefulness of existing ML methods when high precision is requested. Here, we demonstrate that quantum extreme learning machines (QELMs) embody a powerful alternative for witnessing quantum entanglement and, more generally, for estimating features of experimental quantum states. We implement a photonic QELM that leverages the orbital angular momentum of photon pairs as an ancillary degree of freedom to enable informationally complete single-setting measurements of the entanglement shared by their polarization degrees of freedom. Unlike conventional methods, our approach does not require fine-tuning, precise calibration, or refined knowledge of the apparatus. In contrast, it automatically adapts to noise and imprecisions while avoiding overfitting, thus ensuring the robust reconstruction of entanglement witnesses and paving the way to the assessment of quantum features of experimental multi-party states.