In the noisy intermediate-scale quantum (NISQ) regime, quantum devices contain hardware-specific noise sources which restrict device-invariant error mitigation strategies. We explore transfer learning approaches to apply noise models learned on one quantum device to a different device with the help of a small amount of data. We create a real-hardware dataset from two IBM quantum devices, ibm_fez (source) and ibm_marrakesh (target), comprising 170 noisy and ideal circuit output distributions, with device calibration features added. We train a residual neural network on the source device to map noisy to ideal outcomes. The zero-shot transfer test shows a KL divergence of 1.6706 (up from 0.3014), establishing device specificity. With K = 20 fine-tuning samples, KL drops to 1.1924 (28.6% improvement over zero-shot), recovering 34.9% of the gap between zero-shot and in-domain KL. Ablation studies reveal that the major cause of mismatches across devices is CX gate error, followed by readout error. The results show quantum noise can be learned and fine-tuned with minimal samples, and provide a plausible approach to cross-device quantum error mitigation.
Comments
9 pages, 8 figures, 8 tables. Submitted to IEEE Quantum Computing and Engineering (QCE) 2026
In the noisy intermediate-scale quantum (NISQ) regime, quantum devices contain hardware-specific noise sources which restrict device-invariant error mitigation strategies. We explore transfer learning approaches to apply noise models learned on one quantum device to a different device with the help of a small amount of data. We create a real-hardware dataset from two IBM quantum devices, ibm_fez (source) and ibm_marrakesh (target), comprising 170 noisy and ideal circuit output distributions, with device calibration features added. We train a residual neural network on the source device to map noisy to ideal outcomes. The zero-shot transfer test shows a KL divergence of 1.6706 (up from 0.3014), establishing device specificity. With K = 20 fine-tuning samples, KL drops to 1.1924 (28.6% improvement over zero-shot), recovering 34.9% of the gap between zero-shot and in-domain KL. Ablation studies reveal that the major cause of mismatches across devices is CX gate error, followed by readout error. The results show quantum noise can be learned and fine-tuned with minimal samples, and provide a plausible approach to cross-device quantum error mitigation.
Quantum computers, while offering exponential computational advantages, are highly vulnerable to noise-induced errors, particularly during the measurement process. This paper investigates machine learning-based error mitigation strategies with a focus on transfer learning. By initially training a base model on a specific quantum device, we leverage transfer learning to adapt the model for use on different quantum computers. This approach not only achieves superior performance with reduced dataset size and training time but also maintains hardware agnosticism, as demonstrated through tests on various IBM quantum backends. Our method provides a practical solution for enhancing the accuracy and efficiency of noisy intermediate-scale quantum (NISQ) algorithms by improving machine learning-based error mitigation.
We present a systematic investigation of deep learning methods applied to quantum error mitigation of noisy output probability distributions from measured quantum circuits. We compare different architectures, from fully connected neural networks to transformers, and we test different design/training modalities, identifying sequence-to-sequence, attention-based models as the most effective on our datasets. These models consistently produce mitigated distributions that are closer to the ideal outputs when tested on both simulated and real device data obtained from IBM superconducting quantum processing units (QPU) up to five qubits. Across several different circuit depths, our approach outperforms other baseline error mitigation techniques. We perform a series of ablation studies to examine: how different input features (circuit, device properties, noisy output statistics) affect performance; cross-dataset generalization across circuit families; and transfer learning to a different IBM QPU. We observe that generalization performance across similar devices with the same architecture works effectively, without needing to fully retrain models.
Quantum Machine Learning (QML) is an exciting frontier applying the philosophy of quantum computation to classical machine learning for improved learning performance of complicated tasks. However, QML is greatly challenged by realistic issues like noisy quantum measurements and limited labeled quantum data caused by the high cost and technological limitations. This paper explores the novel transfer-learning mechanisms to tackle these issues, which allow the reuse of the knowledge from related quantum tasks for improving model performance and generalization in resource-limited scenarios. Transfer learning, one of the great and successful concepts of classical machine learning, is to transfer the learned parameters or features from the source domain (or task) to the target domain (or task). We further develop and investigate, in quantum, a methodology where parameterized quantum circuits (PQCs) learnt on a source quantum data set have certain parts partially re-used or fine-tuned for a target task, given limited amount of target data. This strategy also avoids large quantum training runs that are not scalable on state-of-the-art noisy intermediate-scale quantum (NISQ) processors. The suggested methods are layerwise quantum adaptation, quantum kernel reuse and hybrid quantum-classical feature transfer. We investigate the robustness of these mechanisms in presence of noise by simulating quantum circuits using depolarizing and readout noise models. Moreover, we show that pretraining of quantum models on synthetic (or simulated) data sets (which are generated using larger, less noisy systems) can be instrumental in learning on real, small-sized quantum devices. Experiment results in quantum classification tasks demonstrate the effectiveness of transfer learning in boosting the accuracy, accelerating the convergence, and lowering the quantum circuit depth of successfully trained model. It is interesting to mention that transfer using layer-wise models became very resistant to quantum noise, as its performance did not degrade significantly for noisy inputs. These results demonstrate that transfer learning can overcome the scarce quantum data and suppress the harmful impact of measurement noise. This work presents a potential venue for enhancing the applicability of QML in the NISQ era. By transferring knowledge from theoretical quantum learning models to actual quantum hardware constraints, transfer learning allows quantum algorithms to learn faster and generalize better across tasks.
Abstract Quantum computing has recently emerged as a novel paradigm for solving problems otherwise intractable on classical computers, yet the practical progress of quantum computers is severely impacted by noise, decoherence, and imperfect gate operations in Noisy Intermediate-Scale Quantum (NISQ) devices. Specific to IBM Quantum hardware, imperfections due to readout error, gate errors, and device drift, which depends on time, resolves limits for the accuracy of outputs from our algorithms. The mitigation strategies currently used, for example, Measurement Error Mitigation (MEM), Zero Noise Extrapolation (ZNE), and Clifford Data Regression (CDR), offer improvement over raw, unmitigated outputs, but they are static and not dynamic; the reduction of noise in within an NISQ device is temporal and can vary in patterns that are non-zero. In addition to this, using the conventional methods provides inconsistent and suboptimal results.This study will provide enhanced reliability with respect to performing quantum computation on IBM Quantum hardware with a unique mitigation framework that is adaptive and computationally efficient for a high-fidelity output. The problem taken on here is the failure of existing error mitigation strategies to provide adaptive noise compensation, as they assume a statically-distributed error, despite the performance-related characteristics already being documented in regards to calibration drift and variation over time.To achieve this, we propose a Hybrid Adaptive Error Mitigation (HAEM) approach that includes baseline measurement error mitigation and an additional correction layer that utilizes realtime calibration information, along with lightweight machine learning models. The approach can be broken down into three steps: (i) perform standard measurement error mitigation to reduce classical readout noise, (ii) execute compact calibration circuits, such as Bell states, GHZ states and Clifford benchmarks to capture then current error profile of the device, and (iii) use a lightweight machine learning model (having been trained on historical and newest calibration data) to dynamically change the mitigation weights applied for the target quantum algorithm. All implementation is done using Qiskit Runtime to ensure low-latency execution, and we completed experiments on both Qiskit Aer noisy simulators and IBM Quantum devices.Initial results suggest HAEM improved the outputs of quantum algorithms' fidelity compared to the unmitigated runs and the baseline measurement error mitigation (MEM) runs. On the noisy simulators, HAEM increased fidelity from a raw performance of 0.65 to 0.87, nearly a 34% improvement. In fairer hardware-like scenarios, fidelity-on-average was persistently 12% better than MEM, with nearly equal time duration for the additional adaptive measurements. The adaptive aspect of HAEM consistently supported subsequent calibration cycles of the device and was effective to continue providing fidelity improvements despite the static mitigation performance degrading.To sum up, the proposed HAEM framework is a modern and viable alternative for error mitigation which brings together the benefits of baseline approaches with corrections that are driven by fixed-function error-corrective, adaptive learning. As both learning-based and standard baseline approaches can only achieve fixed levels of performance, HAEM, through its use of noise mitigation both at execution time and through correlating it with standard data, provides a route towards reliable quantum computations on current IBM hardware. HAEM also presents a contribution to the NISQ error-resilient landscape, and however it is intended that it will continue to improve future applications of hybrid workflows of quantum and classical computation where error mitigation and resilience is important.
Quantum machine learning (QML) models hold promise for leveraging quantum computers in data-driven tasks but their performance on today's noisy quantum hardware remains an open question. In this work, we investigate the impact of shot noise, the statistical noise due to finite sampling in quantum measurements, on the training and performance of quantum neural network (QNN) models. Using a dataset of dielectron collision events from CERN, we train QNNs under varying levels of simulated noise. Three ansatz circuit architectures are explored namely BasicEntanglerLayers (Ansatz-1), SimplifiedTwoDesign (Ansatz-2) and StronglyEntanglingLayers (Ansatz-3). Each model is repeatedly trained for multiple runs at noise levels ranging from zero up to significant noise fractions (0.5), with each training run consisting of 20 epochs of parameter optimization. We evaluate the convergence behavior, final error rates and variability across runs for each noise setting. Our results show a clear degradation of QNN performance as shot noise increases with training loss converging more slowly and to higher values and the model's prediction accuracy deteriorates with larger noise. Notably, the ansatz with fewer parametric gates demonstrates relatively better robustness to noise compared to deeper, more strongly entangling circuits which suffer sharper performance drops under noise. We provide detailed comparisons of the three circuit designs, including training loss curves and statistical summaries of performance at each noise level. These findings shed light on the practical limits of QML models in the Noisy Intermediate-Scale Quantum (NISQ) era and highlight the trade-offs between circuit expressiveness and noise resilience. Finally, we discuss implications for designing noise-aware quantum models and possible strategies to mitigate shot noise effects in quantum learning.
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 H2 molecule in Variational Quantum Eigensolver (VQE) tasks with high fidelity, overcoming the limitations of ZNE in the presence of coherent noise.