We introduce and validate a machine learning-assisted protocol to classify time and space correlations of classical noise acting on a quantum system, using two interacting qubits as probe. We consider different classes of noise, according to their Markovianity and spatial correlations. Leveraging the sensitivity of a coherent population transfer protocol under three distinct driving conditions, the various noises are discriminated by only measuring the final transfer efficiencies. This approach reaches around 90% accuracy with a minimal experimental overhead.
Journal reference
Proceedings of the 3rd International Workshop on AI for Quantum and Quantum for AI (AIQxQIA 2025) co-located with the 28th European Conference on Artificial Intelligence (ECAI 2025)
We introduce and validate a machine learning-assisted protocol to classify time and space correlations of classical noise acting on a quantum system, using two interacting qubits as probe. We consider different classes of noise, according to their Markovianity and spatial correlations. Leveraging the sensitivity of a coherent population transfer protocol under three distinct driving conditions, the various noises are discriminated by only measuring the final transfer efficiencies. This approach reaches around 90% accuracy with a minimal experimental overhead.
We introduce and validate a machine-learning assisted quantum sensing protocol to classify spatial and temporal correlations of classical noise affecting two ultrastrongly coupled qubits. We consider six distinct classes of Markovian and non-Markovian noise. Leveraging the sensitivity of a coherent population transfer protocol under three distinct driving conditions, the various forms of noise are discriminated by only measuring the final transfer efficiencies. Our approach achieves ≳ 94% accuracy in classification providing a near-perfect discrimination between Markovian and non-Markovian noise. The method requires minimal experimental resources, relying on a simple driving scheme providing three inputs to a shallow neural network with no need of measuring time-series data or real-time monitoring. The machine-learning data analysis acquires information from non-idealities of the coherent protocol highlighting how combining these techniques may significantly improve the characterization of quantum-hardware.
Abstract We investigate a machine learning based classification of noise acting on a small quantum network with the aim of detecting spatial or multilevel correlations, and the interplay with Markovianity. We control a three-level system by inducing coherent population transfer exploiting different pulse amplitude combinations as inputs to train a feedforward neural network. We show that supervised learning can classify different types of classical dephasing noise affecting the system. Three non-Markovian (quasi-static correlated, anti-correlated and uncorrelated) and Markovian noises are classified with more than 99% accuracy. On the contrary, correlations of Markovian noise cannot be discriminated with our method. Our approach is robust to statistical measurement errors and retains its effectiveness for physical measurements where only a limited number of samples is available making it very experimental-friendly. Our result paves the way for classifying spatial correlations of noise in quantum architectures.
We propose a hybrid protocol to classify quantum noises using supervised classical machine learning models and simple quantum key distribution protocols. We consider the quantum bit error rates (QBERs) generated in QKD schemes under consideration of different noises, and identify the noise channels with high accuracy for both training and test data. Our protocol classifies quantum noises with high accuracy under the assumption of two different scenarios; in one case we assume two remotely located parties share keys through noisy quantum channels, whereas, in the second case, we simulate the QKD protocols on a gate-based quantum computer, where the gates are afflicted with noise. Alongside efficient classification, our work also throws light on the difference in distribution characteristics of QBERs generated in these two scenarios. Finally, our method is based on classical post processing of data generated from very simplistic quantum protocols, making it readily implementable in the current era of noisy quantum computing with low number of qubits.
Quantum characterization, validation, and verification (QCVV) techniques are used to probe, characterize, diagnose, and detect errors in quantum information processors (QIPs). An important component of any QCVV protocol is a mapping from experimental data to an estimate of a property of a QIP. Machine learning (ML) algorithms can help automate the development of QCVV protocols, creating such maps by learning them from training data. We identify the critical components of "machine-learned" QCVV techniques, and present a rubric for developing them. To demonstrate this approach, we focus on the problem of determining whether noise affecting a single qubit is coherent or stochastic (incoherent) using the data sets originally proposed for gate set tomography. We leverage known ML algorithms to train a classifier distinguishing these two kinds of noise. The accuracy of the classifier depends on how well it can approximate the "natural" geometry of the training data. We find GST data sets generated by a noisy qubit can reliably be separated by linear surfaces, although feature engineering can be necessary. We also show the classifier learned by a support vector machine (SVM) is robust under finite-sample noise.
Correlated noise across multiple qubits poses a significant challenge for achieving scalable and fault-tolerant quantum processors. Despite recent experimental efforts to quantify this noise in various qubit architectures, a comprehensive understanding of its role in qubit dynamics remains elusive. Here, we present an analytical study of the dynamics of driven qubits under spatially correlated noise, including both Markovian and non-Markovian noise. Surprisingly, we find that, while correlated classical noise only leads to correlated decoherence without increasing the quantum coherence in the system, the correlated quantum noise can be exploited to generate entanglement. In particular, we reveal that, in the quantum limit, pure dephasing noise induces a coherent long-range two-qubit Ising interaction that correlates distant qubits. In contrast, for purely transverse noise when qubits are subjected to coherent drives, the correlated quantum noise induces both coherent symmetric exchange and Dzyaloshinskii-Moriya interaction between the qubits, as well as correlated relaxation, both of which give rise to significant entanglement. Remarkably, in this case, we uncover that the system exhibits distinct dynamical phases in different parameter regimes. Finally, we reveal the impact of spatio-temporally correlated 1/f noise on the decoherence rate, and how its temporal correlations restore lost entanglement. Our analysis not only offers critical insights into designing effective error mitigation strategies to reduce harmful effects of correlated noise, but also enables tailored protocols to leverage and harness noise-induced correlations for quantum information processing.
Noisy quantum processors require precise and coherent quantum control to effectively execute quantum algorithms. Key to achieving precise control is the characterization of a quantum device to elucidate important details of environmentally-induced and systematic error sources. Quantum noise spectroscopy (QNS) is one such protocol used to characterize spatiotemporally correlated noise. QNS strives to provide estimates of noise statistical properties by using the quantum system as a dynamical probe. Despite its success, QNS can be challenging as it requires model selection prior to protocol implementation and suffers from scalability issues as system size grows. Here, we leverage machine learning to overcome model selection and assist in learning complex noise environments. In particular, we utilize physics-informed neural networks based on the “gray box” model to learn the dynamics of a two-qubit system subject to spatiotemporally correlated noise. Using training data sets based on dynamical decoupling, we show that the model is capable of learning the underlying power spectral densities for each qubit and the strength of parasitic qubit-qubit interactions. Furthermore, we show that the model can predict state-dependent behavior for multi-qubit observables. The results of this study further display the potential utility of machine learning to assist in the domain of quantum characterization and control.