Entanglement detection with deep neural networks

Entanglement detection is an old and complicated matter. Only for a two-qubit system (or a qubit and a qutrit) it can be solved by the celebrated Peres-Horodecki criteria. Unfortunately, for more complex systems we lack general criteria even if the system is fully characterized. In this talk, we will discuss how we can train neural networks to detect and classify entangled systems from a training set. We will discuss the efficiency of the networks and we will show that they can be used to classify three qubits systems into four possible families.