Two-level systems (TLS) in superconducting quantum devices are a dominant source of decoherence, noise, and device variability, ultimately limiting qubit performance and scalability. Efficient and reliable extraction of TLS parameters from spectroscopy measurements is therefore essential for both device characterization and materials optimization. In this thesis, machine learning methods are developed to accelerate and enhance TLS characterization from superconducting qubit spectroscopy data. The approach combines physics-based simulation of driven qubit–TLS dynamics with supervised learning models trained on synthetic datasets to infer underlying physical parameters from two-dimensional spectroscopy maps. Particular emphasis is placed on understanding how different parameters are encoded in observable features, and on evaluating the extent to which the inverse problem is well-posed. The results demonstrate that machine-learning-assisted inference can significantly reduce the computational cost of parameter estimation compared to conventional fitting-based approaches, while maintaining high accuracy for spectrally encoded parameters such as TLS frequency and coupling strength. In contrast, decoherence parameters are shown to be intrinsically more challenging to extract from short-time spectroscopy data, highlighting fundamental limitations of the measurement protocol rather than the learning model. Overall, this work establishes a data-driven framework for scalable TLS characterization and provides insight into the interplay between physical observability and machine learning-based inference in quantum systems.
Two-level systems (TLS) in superconducting quantum devices are a dominant source of decoherence, noise, and device variability, ultimately limiting qubit performance and scalability. Efficient and reliable extraction of TLS parameters from spectroscopy measurements is therefore essential for both device characterization and materials optimization. In this thesis, machine learning methods are developed to accelerate and enhance TLS characterization from superconducting qubit spectroscopy data. The approach combines physics-based simulation of driven qubit–TLS dynamics with supervised learning models trained on synthetic datasets to infer underlying physical parameters from two-dimensional spectroscopy maps. Particular emphasis is placed on understanding how different parameters are encoded in observable features, and on evaluating the extent to which the inverse problem is well-posed. The results demonstrate that machine-learning-assisted inference can significantly reduce the computational cost of parameter estimation compared to conventional fitting-based approaches, while maintaining high accuracy for spectrally encoded parameters such as TLS frequency and coupling strength. In contrast, decoherence parameters are shown to be intrinsically more challenging to extract from short-time spectroscopy data, highlighting fundamental limitations of the measurement protocol rather than the learning model. Overall, this work establishes a data-driven framework for scalable TLS characterization and provides insight into the interplay between physical observability and machine learning-based inference in quantum systems.
ABSTRACT We introduce a data‐driven approach for extracting two‐level system (TLS) parameters–frequency , coupling strength , dissipation time , and the pure dephasing time , labeled as a 4 component vector , directly from simulated spectroscopy data generated for a single TLS by a form of two‐tone spectroscopy. Specifically, we demonstrate that a custom convolutional neural network model(CNN) can simultaneously predict , , and from the spectroscopy data presented in the form of images. Our results show that the model achieves superior performance to perturbation theory methods in successfully extracting the TLS parameters. Although the model, initially trained on noise‐free data, exhibits a decline in accuracy when evaluated on noisy images, retraining it on a noisy dataset leads to a substantial performance improvement, achieving results comparable to those obtained under noise‐free conditions. Furthermore, the model exhibits higher predictive accuracy for parameters and in comparison to and .
Machine learning (ML) is a promising approach for performing challenging quantum-information tasks such as device characterization, calibration, and control.ML models can train directly on the data produced by a quantum device while remaining agnostic to the quantum nature of the learning task.However, these generic models lack physical interpretability and usually require large datasets in order to learn accurately.Here we incorporate features of quantum mechanics in the design of our ML approach to characterize the dynamics of a quantum device and learn device parameters.This physics-inspired approach outperforms physics-agnostic recurrent neural networks trained on numerically generated and experimental data obtained from continuous weak measurement of a driven superconducting transmon qubit.This demonstration shows how leveraging domain knowledge improves the accuracy and efficiency of this characterization task, thus laying the groundwork for more scalable characterization techniques.
Introduction: The scalability of superconducting quantum processors is fundamentally limited by environmental noise and the exponential measurement overhead required for multi-qubit entanglement verification. Methods: In this work, we introduce a hybrid machine learning (ML) framework utilizing convolutional neural networks (CNNs) to simultaneously characterize decoherence channels and verify multipartite entanglement. The model is trained on synthetic noisy syndrome measurement data augmented with empirical SPAM models and applied to a 5-qubit superconducting device, where the neural network acts as a non-linear entanglement witness. Results: sweeps. Furthermore, it successfully verifies genuine multipartite entanglement with a fidelity prediction of 99.1%, while requiring only 12% of the measurement bases demanded by full quantum state tomography (QST). Discussion: Crucially, our methodology bypasses two distinct challenges: it drastically reduces the measurement complexity inherent in multi-qubit entanglement verification, and it inherently filters out state preparation and measurement (SPAM) errors without requiring noise-amplifying matrix inversions. These numerical results demonstrate that ML-assisted protocols provide a highly resource-efficient pathway for the characterization of noisy intermediate-scale quantum (NISQ) devices.
Microscopic two-level system (TLS) defects remain a primary mechanism of decoherence and operational instability in superconducting transmon qubits, necessitating scalable and automated methods for their characterization. Here, we present and benchmark two complementary analysis pipelines for extracting TLS statistics directly from time-resolved SWAP spectroscopy: one-dimensional decay-rate fitting (1D-DRF), which detects defects via localized enhancements in the qubit relaxation rate, and a deterministic, non-parametric computer-vision framework (2D-CV) that achieves two-dimensional spectral localization by exploiting the temporal persistence of coherent population suppression. We deploy both methods on SWAP spectroscopy measurements from 52 flux-tunable transmon qubits on Rigetti processors with and without moderate (∼10%) post-fabrication frequency trimming via Alternating-Bias Assisted Annealing (ABAA). We show that both pipelines converge on a consistent global characterization of the defect landscape while exhibiting complementary sensitivity across distinct coupling regimes. Crucially, both methods independently reveal a count--loss decoupling under moderate annealing: while the total detectable TLS defect density remains statistically unchanged, the span-integrated dielectric loss is reduced by approximately a factor of two, demonstrating selective suppression of the most strongly dissipative defect channels. These results establish an automated, non-parametric analysis framework for high-throughput hardware diagnostics and provide a statistical baseline for post-fabrication defect engineering in large-scale superconducting quantum processors.
Superconducting qubits are the most prevalent way to design qubits due to their relatively easy manufacturing, tunability and fast gates. However, one of their major problems is the short coherence time, which is influenced by various factors, such as charge noise, flux noise and thermal noise. Several working solutions have been found to mitigate these factors. In contrast, there is no general solution to mitigating the formation of parasitic two-level systems (TLS), which is one of the most dominant causes of error in superconducting circuits. There is no general theory for parasitic TLS, as they form randomly and their physical form varies. Due to this, it is important to be able to characterize parasitic TLSs, so that information can be obtained about their formation and effects on superconducting qubits to improve qubit design. This thesis demonstrates ways to characterize parasitic TLSs in a superconducting qubit. The frequency and the coupling strength of a TLS can be extracted by simulating the spectrum of the system and comparing it to experimental data. The states of a qubit and a TLS are possible to distinguish by utilizing an iSWAP gate to be able to characterize them separately. In this thesis a sample consisting of two fixed-frequency transmon qubits and one tunable transmon coupler was used. The cryostat measurements began by characterizing the sample and calibrating control pulses. After this, data of Rabi oscillations of the system were collected, from which simulations were made and relevant values were obtained through fitting. Then, the iSWAP gate was calibrated and used to swap qubit populations to distinguish the TLS from the qubits. The frequency and the coupling strength of the TLS were obtained from the simulations. At first, the TLS was nearly resonant with the qubit, but then its frequency had moved further away. The separation done with the iSWAP succeeded, which allowed a clear distinction of the separate modes. The results showed that these methods can be utilized to characterize parasitic TLSs. Research can be improved, for example, by measuring TLS coherence times to characterize its properties more accurately.
Quantum computing faces significant challenges from decoherence and noise, which limit the practical implementation of quantum algorithms. While substantial progress has been made in improving individual qubit coherence times, the collective behavior of interconnected qubit systems remains incompletely understood. The connectivity architecture plays a crucial role in determining overall system susceptibility to environmental noise, yet systematic characterization of this relationship has been hindered by computational complexity. We develop a machine learning framework that bridges graph features with quantum device characterization to predict decoherence lifetime directly from connectivity patterns. By representing quantum architectures as connected graphs and using 14 topological features as input to supervised learning models, we achieve accurate lifetime predictions with R2>0.96 for both superconducting and semiconductor platforms. Our analysis reveals fundamentally distinct decoherence mechanisms: superconducting qubits show high sensitivity to global connectivity measures (betweenness centrality δ1=0.484, spectral entropy δ1=0.480), while semiconductor quantum dots exhibit exceptional sensitivity to system scale (node count δ2=0.919, importance = 1.860). The complete failure of cross-platform model transfer (R2 scores of -0.39 and -433.60) emphasizes the platform-specific nature of optimal connectivity design. Our approach enables rapid assessment of quantum architectures without expensive simulations, providing practical guidance for noise-optimized quantum processor design.