Detecting quantum entanglement with unsupervised learning

Quantum properties, such as entanglement and coherence, are indispensable resources in various quantum information processing tasks. However, there still lacks an efficient scalable way to detecting these useful features, especially for high-dimensional multipartite systems. In this work, we exploit the convexity of samples without desired features design unsupervised machine learning method detect presence anomalies. Particularly, context detection, propose a complex-valued neural network composed pseudo-siamese generative adversarial net, then train it with only separable states construct non-linear witnesses entanglement. It is shown via numerical examples, ranging from two-qubit ten-qubit systems, that our able achieve high detection accuracy which above 97.5% on average.Moreover, capable revealing rich structures entanglement, partial among subsystems. Our results readily applicable other Bell nonlocality steerability, thus work could provide powerful tool extract hidden data.