Separability-Entanglement Classifier of 2-qubit State with Machine Learning

Abstract

Quantum entanglement can be used to perform several important tasks such as quantum teleportation and quantum key distribution. In general, however, the problem of determining whether a given quantum state is separable or not is known as NP-hard. In this thesis, we approach the separability problem of a given quantum state with machine learning. Among various supervised learning algorithms, ensemble models based on decision trees such as Random Forest classifier and Gradient Boosting classifier were used for separability problem of a bipartite system. We made an arbitrary two-qubit state labeled with 0 or 1 by PPT (Positive Partial Transpose) criterion. For supervised learning algorithm we prepared the number of 500 separable and entangled states respectively and constructed a dataset for supervised learning algorithm. Also, by adding new feature value obtained from CHA (Convex Hull Approximation) to dataset, we try to improve performances of Random Forest Classifier and Gradient Boosting Classifier. In the case of two-qubit, the result of this research shows that machine learning approach combined with CHA method may work well. Likewise, we can also apply machine learning approach combined with CHA method to two-qutrit separability detecting problem for future works.