Fig. 7: Robustness analysis of the model trained against these noises.
From: Direct entanglement detection of quantum systems using machine learning
The prediction results for the subsystem entropy \(S({\rho }_{A})=-\log (\,\text{tr}\,{\rho }_{A}^{2})\) of the four-qubit test data are shown under control noise (a) and decoherence noise (b). Each control noise strength ξ and each decoherence strength γ is tested with M = 500 samples. The average error, defined as the average distance between the actual entropy and the predicted entropy, is calculated as \(\epsilon =\frac{1}{3M}\mathop{\sum }\nolimits_{m = 1}^{M}\mathop{\sum }\nolimits_{{\rho }_{A} = [1]}^{[123]}| {S}_{m}^{{\rm{th}}}({\rho }_{A})-{S}_{m}^{{\rm{ML}}}({\rho }_{A})|\).
