An experiment is performed to reconstruct an unknown photonic quantum state with a limited amount of copies. A semiquantum reinforcement learning approach is employed to adapt one qubit state, an “agent,” to an unknown quantum state, an “environment,” by successive single‐shot measurements and feedback, in order to achieve maximum overlap. The experimental learning device herein, composed of a quantum photonics setup, can adjust the corresponding parameters to rotate the agent system based on the measurement outcomes “0” or “1” in the environment (i.e., reward/punishment signals). The results show that, when assisted by such a quantum machine learning technique, fidelities of the deterministic single‐photon agent states can achieve over 88% under a proper reward/punishment ratio within 50 iterations. This protocol offers a tool for reconstructing an unknown quantum state when only limited copies are provided, and can also be extended to higher dimensions, multipartite, and mixed quantum state scenarios.

Quantum coherence is the most distinguished feature of quantum mechanics. It lies at the heart of the quantum-information technologies as the fundamental resource and is also related to other quantum resources, including entanglement. It plays a critical role in various fields, even in biology. Nevertheless, the rigorous and systematic resource-theoretic framework of coherence has just been developed recently, and several coherence measures are proposed. Experimentally, the usual method to measure coherence is to perform state tomography and use mathematical expressions. Here, we alternatively develop a method to measure coherence directly using its most essential behavior-the interference fringes. The ancilla states are mixed into the target state with various ratios, and the minimal ratio that makes the interference fringes of the "mixed state" vanish is taken as the quantity of coherence. We also use the witness observable to witness coherence, and the optimal witness constitutes another direct method to measure coherence. For comparison, we perform tomography and calculate l_{1} norm of coherence, which coincides with the results of the other two methods in our situation. Our methods are explicit and robust, providing a nice alternative to the tomographic technique.

The spectral theorem of von Neumann has been widely applied in various areas, such as the characteristic spectral lines of atoms. It has been recently proposed that dynamical evolution also possesses spectral lines. As the most intrinsic property of evolution, the behavior of these spectra can, in principle, exhibit almost every feature of this evolution, among which the most attractive topic is non-Markovianity, i.e., the memory effects during evolution. Here, we develop a method to detect these spectra, and moreover, we experimentally examine the relation between the spectral behavior and non-Markovianity by engineering the environment to prepare dynamical maps with different non-Markovian properties and then detecting the dynamical behavior of the spectral values. These spectra will lead to a witness for essential non-Markovianity. We also experimentally verify another simplified witness method for essential non-Markovianity. Interestingly, in both cases, we observe the sudden transition from essential non-Markovianity to something else. Our work shows the role of the spectra of evolution in the studies of non-Makovianity and provides the alternative methods to characterize non-Markovian behavior.