arXiv:2609.22768 · Frontiers of Physics
Neural network-based multipartite entanglement classification with prior guidance from quantum uncertainty relations
Qiyi Li, Xiao Zheng, Guofeng Zhang
Abstract
Quantum entanglement is a crucial resource in quantum information processing, yet its efficient, scalable and robust classification in multipartite systems remains theoretically challenging. Although supervised machinelearning has been applied to this task, most existing methods still suffer from high measurement costs, computational consumption, and weak noise robustness. In this work, by incorporating multipartite uncertainty relations as prior guidance, we propose a neural network approach to classify distinct Stochastic Local Operations and Classical Communication (SLOCC) multipartite entanglement classes based on states sampled from their local unitary (LU) orbits. Compared with traditional techniques, our method reduces experimental measurement-resource requirements and computational overhead, showing high adaptability to large-scale systems. The classification accuracy of our method reaches 99.5% in 20-qubit systems. The numerical validation is performed on states generated by random LU transformations, which preserve the SLOCC class. Within this setting, the proposed method offers strong effectiveness, scalability, and robustness for multipartite entanglement classification.
AI Summary
The paper investigates whether uncertainty-based features can reduce the input and measurement requirements of supervised multipartite entanglement classification. It uses conditional variances motivated by a quantum-control-assisted multipartite uncertainty relation to construct an $n$-component feature vector from local Pauli $X$ and $Z$ measurement statistics, rather than supplying a full density matrix to a neural network. A standard multilayer perceptron is trained to distinguish states generated from GHZ, W, and linear-cluster representatives. The evaluation uses classical numerical simulations of random local-unitary (LU) transformations of these representatives, with an 80:20 training/test split. The reported dataset contains 1,890,000 states, including pure states, mixtures of two pure states, and states with added white noise. For 10 qubits, three-way classification reaches 99.8% accuracy for pure states and 97.3% for mixtures. Binary GHZ-versus-W classification reaches 99.5% and 98.5%, respectively. GHZ-versus-linear-cluster classification exceeds 99% for 11–20 qubits, reaching approximately 99.5% at 20 qubits. Noise tests use $\rho_{\mathrm{noisy}}=p\rho+(1-p)I/d$ with $p=0.7$–$0.9$. The stated limitations include lower accuracy for mixed states, the absence of experimental validation, and restriction to LU-generated datasets. Although LU transformations preserve SLOCC class, these datasets do not explore general non-unitary SLOCC transformations. Thus, the reported accuracies concern discrimination among the sampled LU orbits and their constructed mixtures or noisy variants, rather than validation across entire SLOCC classes.
Central claim may not hold: The introduction overstates the benchmark scope: the 20-qubit result concerns only pure-state binary GHZ-versus-linear-cluster classification, not ternary, mixed-state or noisy-state classification at that size. The claimed polynomial measurement and sample-generation requirements are not established by the linear feature dimension, because the cost of estimating the conditional variances is not analyzed. The mixed-state labels lack an entanglement-class justification, since mixtures of LU-equivalent pure states can be fully separable. The three-qubit GHZ and linear cluster states are LU-equivalent, so their reported discrimination cannot represent classification of distinct SLOCC entanglement classes.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 10 pages,6,figures,Front.Phys(2026)in press