Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

IEEE International Conference on Cybernetics (CYBCONF)

Physics-Informed Multi-Head Networks for Adaptive Quantum Noise Reconstruction and Mitigation

Qixiang Zhang, Zigui Zhang, Yang He, Zibo Miao

Published Apr 1, 2026• 0 citations
Publisher

Abstract

Quantum error mitigation (QEM) is a prerequisite for extracting utility from noisy intermediate-scale quantum (NISQ) processors. Among emerging protocols, tensor network error mitigation (TEM) offers scalable noise inversion, yet its efficacy is strictly bounded by the fidelity of the underlying noise channel reconstruction. The prevailing “cold-start” unsupervised learning paradigm typically treats noise characterization as a black-box optimization problem, which may result in a nonconvex landscape, where standard optimizers consistently stall at suboptimal local minima. In this paper, we propose a physicsinformed quantum multi-head network (QMHN) to overcome the barrier. By embedding the structure of the sparse Pauli-Lindblad model (SPLM) directly into the neural architecture, our method explicitly decouples the learning manifolds of local single-qubit dissipation and correlated two-qubit crosstalk errors. Crucially, we demonstrate that such physical inductive bias reduces the training convergence time by approximately 88% and significantly reduces the reconstruction error floor compared to monolithic direct parameter optimization. To further approach the theoretical fidelity limit, we introduce an adaptive prior-guided mechanism, which dynamically balances simulated measurement statistics with prior physical constraints. Validated against a controlled ideal prior, our framework navigates the complex optimization landscape to achieve a reconstruction precision of 10−6. When applied to TEM on deep quantum circuits, this ultra-high-fidelity model enables near-exact observable recovery where conventional characterization methods fail. Our results demonstrate that incorporating physical structural constraints into tensor-network learning is essential for scalable, highprecision quantum noise characterization.

DOI
10.1109/cybconf69298.2026.11670278

Related papers

Scalable Quantum Error Mitigation with Physically Informed Graph Neural Networks

Huaxin Wang, Xinge Wu, Jiajun Liu +4 more

Quantum error mitigation (QEM) provides a practical route for estimating reliable observables on noisy intermediate-scale quantum (NISQ) devices. Traditional QEM strategies, including zero-noise extrapolation (ZNE) and Clifford data regression (CDR), rely on noise scaling or global regression, and their performance is constrained by the exponential growth of the system degrees of freedom. We construct a graph-enhanced mitigation (GEM) framework, which incorporates physical information into the model representation. In this work, quantum circuits are encoded as attributed graphs. Hardware-level physical information is mapped to node and edge features: local noise parameters such as calibration parameters $T_1$, $T_2$, and readout errors are encoded at nodes, while coupling-related information such as two-qubit gate errors is encoded as edge features. Graph neural networks are used to model how errors propagate along the physical coupling structure and build up into non-local correlations. This allows the model to capture local interactions and part of the resulting non-local correlations across qubits. A dual-branch affine correction is applied to maintain consistency with physical constraints. Experiments on 10-qubit and 16-qubit random circuits executed on superconducting quantum processors show that GEM provides a level of accuracy comparable to CDR at small scales, while yielding lower mean absolute error and improved stability in zero-shot transfer to larger systems. Results of the traditional QEM strategy indicate that global regression methods remain effective in low-dimensional settings but become less reliable as system degrees of freedom grow. In contrast, GEM makes use of local physical structures to show better scalability and generalization, while preserving the overall error propagation patterns. This work provides a practical scalable approach to QEM for NISQ devices.

quant-pharXiv:2604.16815•Quantum Machine Intelligence•Apr 21, 2026•1 citation

Scalable tensor-network error mitigation for near-term quantum computing

Filippov, Sergei, Matea Leahy, Matteo A. C. Rossi +1 more

Until fault-tolerance becomes implementable at scale, quantum computing will heavily rely on noise mitigation techniques. While methods such as zero noise extrapolation with probabilistic error amplification (ZNE-PEA) and probabilistic error cancellation (PEC) have been successfully tested on hardware recently, their scalability to larger circuits may be limited. Here, we introduce the tensor-network error mitigation (TEM) algorithm, which acts in post-processing to correct the noise-induced errors in estimations of physical observables. The method consists of the construction of a tensor network representing the inverse of the global noise channel affecting the state of the quantum processor, and the consequent application of the map to informationally complete measurement outcomes obtained from the noisy state. TEM does therefore not require additional quantum operations other than the implementation of informationally complete POVMs, which can be achieved through randomised local measurements. The key advantage of TEM is that the measurement overhead is quadratically smaller than in PEC. We test TEM extensively in numerical simulations in different regimes. We find that TEM can be applied to circuits of twice the depth compared to what is achievable with PEC under realistic conditions with sparse Pauli-Lindblad noise, such as those in [E. van den Berg et al., Nat. Phys. (2023)]. By using Clifford circuits, we explore the capabilities of the method in wider and deeper circuits with lower noise levels. We find that in the case of 100 qubits and depth 100, both PEC and ZNE fail to produce accurate results by using $\sim 10^5$ shots, while TEM succeeds.

arXiv:2307.11740•Jul 21, 2023•93 citations

Accelerating Noisy Variational Quantum Algorithms with Physics-Informed Denoising Networks

Jie Liu, Xin Wang

Variational quantum algorithms are promising for near-term quantum computing, but are severely limited by hardware noise and the substantial circuit overhead required for error mitigation methods such as Zero-Noise Extrapolation (ZNE). We propose a Physics-Informed Denoising Network (PIDN) that reduces the cost of ZNE by learning a surrogate model of its optimization dynamics. By viewing the variational update as a trajectory in the parameter space, PIDN is trained to reproduce ZNE-mitigated expectation values and gradient directions while incorporating a physics-informed loss that preserves the gradient descent dynamics. Once trained, PIDN replaces repeated multi-noise evaluations with denoised expectation and gradient estimation directly from the current noisy observation and the historical trajectory, significantly reducing circuit executions. We benchmark the approach on the quantum approximate optimization algorithm for 3-regular graphs, Sherrington-Kirkpatrick, and transverse-field Ising models, as well as the variational quantum eigensolver for LiH, BeH$_2$ and H$_2$O. Across all tasks, PIDN attains performance comparable to ZNE, while reducing the number of circuit executions by a factor of approximately 4 to 6. Gradient cosine similarity with ZNE remains above 0.95 throughout training. Robustness analysis shows that PIDN fails only when ZNE itself becomes unreliable, and ablation studies confirm the necessity of the physics-informed loss for maintaining directional consistency. We further find that PIDN tracks optimization dynamics most accurately when the effective loss landscape retains strong low-frequency structure. These results establish PIDN as a scalable, resource-efficient strategy for noise-resilient variational optimization in the noisy intermediate-scale quantum regime.

quant-pharXiv:2605.02066•Physical Review Research•May 5, 2026•0 citations

Calibration-Aware Graph Neural Networks for Robust and Scalable Quantum Error Mitigation

Zhihao Shen

Abstract Noisy Intermediate-Scale Quantum (NISQ) computing is fundamentally constrained by hardware noise that is not only large but also temporally drifting and structurally complex. Traditional quantum error mitigation (QEM) techniques, such as Zero-Noise Extrapolation (ZNE), often rely on simplified noise assumptions that degrade under coherent errors and deep circuit structures. Conversely, existing machine learning-based QEM approaches typically treat device noise as a static background, leading to poor generalization when hardware calibration shifts. In this work, we propose \textbf{Calib-GNN}, a novel framework that explicitly integrates real-time hardware calibration data into a Graph Neural Network (GNN) architecture. By embedding gate-specific error rates directly into the graph representation, our model learns a dynamic mapping from noisy measurements to ideal expectation values. Through extensive simulations based on real noise profiles from the IBM Quantum \texttt{ibm\_fez} backend, we demonstrate that Calib-GNN significantly outperforms standard baselines. Specifically, it reduces the Root Mean Square Error (RMSE) by approximately 14\% compared to ZNE on random circuits and exhibits superior robustness against severe noise drift where traditional models fail. Furthermore, we show that Calib-GNN effectively scales to deeper circuits and recovers the ground state energy of the $H_2$ molecule in Variational Quantum Eigensolver (VQE) tasks with high fidelity, overcoming the limitations of ZNE in the presence of coherent noise.

•Dec 2, 2025•0 citations

Learning Spatiotemporally Correlated Noise in Multi-Qubit Systems with Neural Networks

Abhiram Nallamalli, Shantanu Misra, Gregory Quiroz

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.

IEEE International Conference on Quantum Computing and Engineering (QCE)•Sep 15, 2024•0 citations

Exact learning of quantum noise with tensor networks

Nicola Pancotti, Vedika Saravanan, Krysta Svore

Accurate noise models are essential for high-performance quantum error correction, yet characterizing the noise of a quantum device typically requires dedicated experiments. We present a variational framework that learns the noise model directly from quantum-error-correction syndrome and logical-observable data collected during error-corrected memory experiments. The fault-event probabilities are treated as variational parameters and optimized via gradient descent to minimize the binary cross-entropy between the decoder's predictions and experimental logical-observable outcomes. We prove that this objective is principled rather than ad hoc: a sufficiently expressive noise ansatz attains the information-theoretic minimum logical error rate. We instantiate this framework using a tensor-network decoder, which provides exact maximum-likelihood decoding and analytically differentiable gradients with respect to all noise parameters. Using circuit-level data from Google's Sycamore processor, and starting from an uninformed prior, the optimization recovers noise models whose logical error rates agree to within $2\%$ with those of Google's independently characterized detector error model. The mean squared error between the learned and reference noise parameters shows a clear overall decrease throughout training, confirming that the method recovers physically meaningful noise structure, not merely parameters that happen to decode well. We further demonstrate that the optimization can track device drifts in real time via warm-started updates, maintaining near-optimal decoding performance under synthetically evolving noise without the need for re-characterization. The approach is decoder-agnostic in its formulation and naturally extends to correlated noise models.

quant-pharXiv:2609.00169•Sep 2, 2026•3 citations
Quantum error mitigation (QEM) is a prerequisite for extracting utility from noisy intermediate-scale quantum (NISQ) processors. Among emerging protocols, tensor network error mitigation (TEM) offers scalable noise inversion, yet its efficacy is strictly bounded by the fidelity of the underlying noise channel reconstruction. The prevailing “cold-start” unsupervised learning paradigm typically treats noise characterization as a black-box optimization problem, which may result in a nonconvex landscape, where standard optimizers consistently stall at suboptimal local minima. In this paper, we propose a physicsinformed quantum multi-head network (QMHN) to overcome the barrier. By embedding the structure of the sparse Pauli-Lindblad model (SPLM) directly into the neural architecture, our method explicitly decouples the learning manifolds of local single-qubit dissipation and correlated two-qubit crosstalk errors. Crucially, we demonstrate that such physical inductive bias reduces the training convergence time by approximately 88% and significantly reduces the reconstruction error floor compared to monolithic direct parameter optimization. To further approach the theoretical fidelity limit, we introduce an adaptive prior-guided mechanism, which dynamically balances simulated measurement statistics with prior physical constraints. Validated against a controlled ideal prior, our framework navigates the complex optimization landscape to achieve a reconstruction precision of 10−6. When applied to TEM on deep quantum circuits, this ultra-high-fidelity model enables near-exact observable recovery where conventional characterization methods fail. Our results demonstrate that incorporating physical structural constraints into tensor-network learning is essential for scalable, highprecision quantum noise characterization.
Quantum error mitigation (QEM) provides a practical route for estimating reliable observables on noisy intermediate-scale quantum (NISQ) devices. Traditional QEM strategies, including zero-noise extrapolation (ZNE) and Clifford data regression (CDR), rely on noise scaling or global regression, and their performance is constrained by the exponential growth of the system degrees of freedom. We construct a graph-enhanced mitigation (GEM) framework, which incorporates physical information into the model representation. In this work, quantum circuits are encoded as attributed graphs. Hardware-level physical information is mapped to node and edge features: local noise parameters such as calibration parameters T1T_1T1​, T2T_2T2​, and readout errors are encoded at nodes, while coupling-related information such as two-qubit gate errors is encoded as edge features. Graph neural networks are used to model how errors propagate along the physical coupling structure and build up into non-local correlations. This allows the model to capture local interactions and part of the resulting non-local correlations across qubits. A dual-branch affine correction is applied to maintain consistency with physical constraints. Experiments on 10-qubit and 16-qubit random circuits executed on superconducting quantum processors show that GEM provides a level of accuracy comparable to CDR at small scales, while yielding lower mean absolute error and improved stability in zero-shot transfer to larger systems. Results of the traditional QEM strategy indicate that global regression methods remain effective in low-dimensional settings but become less reliable as system degrees of freedom grow. In contrast, GEM makes use of local physical structures to show better scalability and generalization, while preserving the overall error propagation patterns. This work provides a practical scalable approach to QEM for NISQ devices.
Until fault-tolerance becomes implementable at scale, quantum computing will heavily rely on noise mitigation techniques. While methods such as zero noise extrapolation with probabilistic error amplification (ZNE-PEA) and probabilistic error cancellation (PEC) have been successfully tested on hardware recently, their scalability to larger circuits may be limited. Here, we introduce the tensor-network error mitigation (TEM) algorithm, which acts in post-processing to correct the noise-induced errors in estimations of physical observables. The method consists of the construction of a tensor network representing the inverse of the global noise channel affecting the state of the quantum processor, and the consequent application of the map to informationally complete measurement outcomes obtained from the noisy state. TEM does therefore not require additional quantum operations other than the implementation of informationally complete POVMs, which can be achieved through randomised local measurements. The key advantage of TEM is that the measurement overhead is quadratically smaller than in PEC. We test TEM extensively in numerical simulations in different regimes. We find that TEM can be applied to circuits of twice the depth compared to what is achievable with PEC under realistic conditions with sparse Pauli-Lindblad noise, such as those in [E. van den Berg et al., Nat. Phys. (2023)]. By using Clifford circuits, we explore the capabilities of the method in wider and deeper circuits with lower noise levels. We find that in the case of 100 qubits and depth 100, both PEC and ZNE fail to produce accurate results by using ∼105\sim 10^5∼105 shots, while TEM succeeds.
Variational quantum algorithms are promising for near-term quantum computing, but are severely limited by hardware noise and the substantial circuit overhead required for error mitigation methods such as Zero-Noise Extrapolation (ZNE). We propose a Physics-Informed Denoising Network (PIDN) that reduces the cost of ZNE by learning a surrogate model of its optimization dynamics. By viewing the variational update as a trajectory in the parameter space, PIDN is trained to reproduce ZNE-mitigated expectation values and gradient directions while incorporating a physics-informed loss that preserves the gradient descent dynamics. Once trained, PIDN replaces repeated multi-noise evaluations with denoised expectation and gradient estimation directly from the current noisy observation and the historical trajectory, significantly reducing circuit executions. We benchmark the approach on the quantum approximate optimization algorithm for 3-regular graphs, Sherrington-Kirkpatrick, and transverse-field Ising models, as well as the variational quantum eigensolver for LiH, BeH2_22​ and H2_22​O. Across all tasks, PIDN attains performance comparable to ZNE, while reducing the number of circuit executions by a factor of approximately 4 to 6. Gradient cosine similarity with ZNE remains above 0.95 throughout training. Robustness analysis shows that PIDN fails only when ZNE itself becomes unreliable, and ablation studies confirm the necessity of the physics-informed loss for maintaining directional consistency. We further find that PIDN tracks optimization dynamics most accurately when the effective loss landscape retains strong low-frequency structure. These results establish PIDN as a scalable, resource-efficient strategy for noise-resilient variational optimization in the noisy intermediate-scale quantum regime.
Abstract Noisy Intermediate-Scale Quantum (NISQ) computing is fundamentally constrained by hardware noise that is not only large but also temporally drifting and structurally complex. Traditional quantum error mitigation (QEM) techniques, such as Zero-Noise Extrapolation (ZNE), often rely on simplified noise assumptions that degrade under coherent errors and deep circuit structures. Conversely, existing machine learning-based QEM approaches typically treat device noise as a static background, leading to poor generalization when hardware calibration shifts. In this work, we propose \textbf{Calib-GNN}, a novel framework that explicitly integrates real-time hardware calibration data into a Graph Neural Network (GNN) architecture. By embedding gate-specific error rates directly into the graph representation, our model learns a dynamic mapping from noisy measurements to ideal expectation values. Through extensive simulations based on real noise profiles from the IBM Quantum \texttt{ibm\_fez} backend, we demonstrate that Calib-GNN significantly outperforms standard baselines. Specifically, it reduces the Root Mean Square Error (RMSE) by approximately 14\% compared to ZNE on random circuits and exhibits superior robustness against severe noise drift where traditional models fail. Furthermore, we show that Calib-GNN effectively scales to deeper circuits and recovers the ground state energy of the H2H_2H2​ molecule in Variational Quantum Eigensolver (VQE) tasks with high fidelity, overcoming the limitations of ZNE in the presence of coherent noise.
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.
Accurate noise models are essential for high-performance quantum error correction, yet characterizing the noise of a quantum device typically requires dedicated experiments. We present a variational framework that learns the noise model directly from quantum-error-correction syndrome and logical-observable data collected during error-corrected memory experiments. The fault-event probabilities are treated as variational parameters and optimized via gradient descent to minimize the binary cross-entropy between the decoder's predictions and experimental logical-observable outcomes. We prove that this objective is principled rather than ad hoc: a sufficiently expressive noise ansatz attains the information-theoretic minimum logical error rate. We instantiate this framework using a tensor-network decoder, which provides exact maximum-likelihood decoding and analytically differentiable gradients with respect to all noise parameters. Using circuit-level data from Google's Sycamore processor, and starting from an uninformed prior, the optimization recovers noise models whose logical error rates agree to within 2%2\%2% with those of Google's independently characterized detector error model. The mean squared error between the learned and reference noise parameters shows a clear overall decrease throughout training, confirming that the method recovers physically meaningful noise structure, not merely parameters that happen to decode well. We further demonstrate that the optimization can track device drifts in real time via warm-started updates, maintaining near-optimal decoding performance under synthetically evolving noise without the need for re-characterization. The approach is decoder-agnostic in its formulation and naturally extends to correlated noise models.