Abstract:Entanglement detection in high dimensional systems is a NP-hard problem since it is lacking an efficient way. Given a bipartite quantum state of interest free entanglement can be detected efficiently by the PPT-criterion (Peres-Horodecki criterion), in contrast to detecting bound entanglement, i.e. a curious form of entanglement that can also not be distilled into maximally (free) entangled states. Only a few bound entangled states have been found, typically by constructing dedicated entanglement witnesses, so naturally the question arises how large is the volume of those states. We define a large family of magically symmetric states of bipartite qutrits for which we find $82\%$ to be free entangled, $2\%$ to be certainly separable and as much as $10\%$ to be bound entangled, which shows that this kind of entanglement is not rare. Via various machine learning algorithms we can confirm that the remaining $6\%$ of states are more likely to belonging to the set of separable states than bound entangled states. Most important we find via dimension reduction algorithms that there is a strong $2$-dimensional (linear) sub-structure in the set of bound entangled states. This revealed structure opens a novel path to find and characterize bound entanglement towards solving the long-standing problem of what the existence of bound entanglement is implying.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the difficulty of detecting quantum entanglement (especially boundary entanglement) in high - dimensional systems. Specifically: 1. **Background problems**: - In high - dimensional quantum systems, detecting entanglement is an NP - hard problem due to the lack of effective detection methods. - For bipartite quantum states, free entanglement can be efficiently detected by the PPT criterion (Peres - Horodecki criterion), but it is very difficult to detect boundary entanglement. - Currently, there are few known boundary - entangled states, and they are usually discovered by constructing specific entanglement witnesses. 2. **Research objectives**: - The researchers defined a class of bipartite qutrits with magic symmetry and detected the entanglement properties of these states by machine - learning algorithms. - Specific objectives include: - Detecting the proportions of free entanglement, boundary entanglement, and separable states. - Revealing the structural characteristics of boundary - entangled states through dimension - reduction algorithms. - Evaluating whether an unknown state (a state whose entanglement property cannot be determined by existing methods) is more likely to be a separable state or a boundary - entangled state. 3. **Main contributions**: - It was found that 10% of the states are boundary - entangled states, indicating that boundary entanglement is not rare in these systems. - Confirmed by machine - learning algorithms, the remaining 6% of the unknown states are more likely to be separable states. - Revealed the strong structural characteristics of boundary - entangled states in two - dimensional linear subspaces, providing a new path for future research. 4. **Methods**: - Classify states using machine - learning algorithms (such as the nearest - neighbor algorithm and the gradient - boosted tree). - Reveal the structural characteristics of boundary - entangled states through dimension - reduction algorithms (such as principal component analysis). - Generate a large number of entanglement witnesses and apply these witnesses to detect boundary - entangled states. 5. **Results**: - Among the studied states, 70% are free - entangled states, 17% are boundary - entangled states, 3% are separable states, and 10% are still unknown states. - Through machine - learning algorithms, it was confirmed that the unknown states are more likely to be separable states. - Revealed the strong structural characteristics of boundary - entangled states in two - dimensional linear subspaces, providing new ideas for future detection. In general, by combining the methods of machine learning and quantum information theory, this paper successfully addresses some of the challenges of the NP - hard problem of detecting boundary entanglement in high - dimensional systems and provides new directions for future research.

Free versus Bound Entanglement: Machine learning tackling a NP-hard problem

Optimized Detection of High-Dimensional Entanglement

Large-Scale Quantum Separability Through a Reproducible Machine Learning Lens

Separability-entanglement Classifier Via Machine Learning

Entanglement Structure Detection via Machine Learning

Bipartite Bound Entanglement

Detecting Entanglement by Pure Bosonic Extension

Almost complete solution for the NP-hard separability problem of Bell diagonal qutrits

Detecting quantum entanglement with unsupervised learning

Detecting the dimensionality of genuine multiparticle entanglement

Detecting the dimensionality of genuine multi-particle entanglement

Quantifying Quantum Entanglement Via Machine Learning Models

Data-driven approach to mixed-state multipartite entanglement characterisation

High Schmidt Number Concentration in Quantum Bound Entangled States

A nonlinear criterion for characterizing high-dimensional multipartite entanglement

Entanglement Detection by Approximate Entanglement Witnesses

Detecting the Concurrence of an Unknown State with a Single Observable

Machine Learning Entanglement Freedom Or: How I Learned to Stop Worrying and Love Linear Regression

Bounds and guarantees for learning and entanglement

Entanglement-Based Machine Learning On A Quantum Computer