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

arXiv:1907.11947 · Machine Learning Science and Technology

Repetitive readout enhanced by machine learning

Genyue Liu, Mo Chen, Yi-Xiang Liu, David Layden, Paola Cappellaro

Published Feb 4, 2020• 30 citations
arXivHTMLPDF

Abstract

Abstract Single-shot readout is a key component for scalable quantum information processing. However, many solid-state qubits with favorable properties lack the single-shot readout capability. One solution is to use the repetitive quantum-non-demolition readout technique, where the qubit is correlated with an ancilla, which is subsequently read out. The readout fidelity is therefore limited by the back-action on the qubit from the measurement. Traditionally, a threshold method is taken, where only the total photon count is used to discriminate qubit state, discarding all the information of the back-action hidden in the time trace of repetitive readout measurement. Here we show by using machine learning (ML), one obtains higher readout fidelity by taking advantage of the time trace data. ML is able to identify when back-action happened, and correctly read out the original state. Since the information is already recorded (but usually discarded), this improvement in fidelity does not consume additional experimental time, and could be directly applied to preparation-by-measurement and quantum metrology applications involving repetitive readout.

DOI
10.1088/2632-2153/ab4e24

Related papers

Machine-learning-assisted electron-spin readout of nitrogen-vacancy center in diamond

Peng Qian, Xue Lin, Feifei Zhou +5 more

Machine learning is a powerful tool in finding hidden data patterns for quantum information processing. Here, we introduce this method into the optical readout of electron-spin states in diamond via single-photon collection and demonstrate improved readout precision at room temperature. The traditional method of summing photon counts in a time gate loses all the timing information crudely. We find that changing the gate width can only optimize the contrast or the state variance, not both. In comparison, machine learning adaptively learns from time-resolved fluorescence data and offers the optimal data processing model that elaborately weights each time bin to maximize the extracted information. It is shown that our method can repair the processing result from imperfect data, reducing 7% in spin readout error while optimizing the contrast. Note that these improvements only involve recording photon time traces and consume no additional experimental time, and they are, thus, robust and free. Our machine learning method implies a wide range of applications in the precision measurement and optical detection of states.

arXiv:2102.08558•Applied Physics Letters•Feb 22, 2021•32 citations

Quantum non-demolition readout of an electron spin in silicon

Jun Yoneda, Kenta Takeda, Akito Noiri +5 more

While single-shot detection of silicon spin qubits is now a laboratory routine, the need for quantum error correction in a large-scale quantum computing device demands a quantum non-demolition (QND) implementation. Unlike conventional counterparts, the QND spin readout imposes minimal disturbance to the probed spin polarization and can therefore be repeated to extinguish measurement errors. Here, we show that an electron spin qubit in silicon can be measured in a highly non-demolition manner by probing another electron spin in a neighboring dot Ising-coupled to the qubit spin. The high non-demolition fidelity (99% on average) enables over 20 readout repetitions of a single spin state, yielding an overall average measurement fidelity of up to 95% within 1.2 ms. We further demonstrate that our repetitive QND readout protocol can realize heralded high-fidelity (>99.6%) ground-state preparation. Our QND-based measurement and preparation, mediated by a second qubit of the same kind, will allow for a wide class of quantum information protocols with electron spins in silicon without compromising the architectural homogeneity.

arXiv:1910.11963•Nature Communications•Mar 2, 2020•80 citations

Time-series based quantum state discrimination

Samuel Jung, Neel Vora, Akel Hashim +2 more

Accurate quantum state readout is crucial for error correction and algorithms, but measurement errors are detrimental. Readout fidelity is typically limited by a poor signal-to-noise ratio (SNR) and energy relaxation ($T_1$ decay), a significant problem for superconducting qubits. While most approaches classify results using clustering algorithms on integrated readout signals, these methods cannot distinguish a qubit that was initially in the ground state from one that decayed to it during measurement. We instead propose using machine learning (ML) on the raw, non-integrated analog signal. We apply time-series classification models, such as a long short-term memory (LSTM) network, to the full data trajectory. We find that our LSTM model, combined with filtering and feature engineering, consistently outperforms clustering. The largest improvements come from reclassifying points in the boundary regions between clusters. These points correspond to atypical measurement records, likely due to transient or noisy features lost during data integration. By retaining temporal information, sequence-aware models like LSTMs can better discriminate these trajectories, whereas clustering methods based on integrated values are more prone to misclassification.

quant-pharXiv:2601.19057•Jan 28, 2026•1 citation

Optimal repetitive readout of single solid-state spins determined by Fisher information

Zhiyuan Zhao, Shaoyi Xu, Qian Shi +10 more

Quantum systems, including superconducting circuits, trapped ions, quantum dots, solid-state defects, etc., have achieved considerable advancements in readout fidelity. However, the widely used threshold method disregards the importance of temporal characteristics of the signal during continuous measurements, leading to information loss. Here, we applied Fisher information to quantify the potential enhancement by using the temporal information. We proposed an optimal data processing method and derived the condition for equality to the error constrained by Fisher information. Applying this method to the single-shot readout of a [Formula: see text] nuclear spin of nitrogen-vacancy center marks a 33.8 ± 1.2% error reduction compared to the threshold method, improving the fidelity to 99.649(5)%. The versatility of our method is further demonstrated through the single-shot readout of charge states. This work illustrates the advantage of using photon arrival time for improving readout fidelity without upgrading experimental hardware. The method of using Fisher information to analyze readout fidelity also holds promise for broad applicability to other systems.

Science Advances•Dec 6, 2024•1 citation

Singlet-triplet-state readout in silicon metal-oxide-semiconductor double quantum dots

Rong-Long Ma, Sheng-Kai Zhu, Zhenzhen Kong +9 more

High-fidelity singlet-triplet state readout is essential for large-scale quantum computing. However, the widely used threshold method of comparing a mean value with the fixed threshold will limit the judgment accuracy, especially for the relaxed triplet state, under the restriction of relaxation time and signal-to-noise ratio. Here, we achieve an enhanced latching readout based on Pauli spin blockade in a Si-MOS double quantum dot device and demonstrate an average singlet-triplet-state readout fidelity of 97.59% by the threshold method. We reveal the inherent deficiency of the threshold method for the relaxed triplet-state classification and introduce machine learning as a noise-resilient and relaxation-independent readout method to reduce the misjudgment. The readout fidelity for classifying the simulated single-shot traces can be improved to 99.67% by the machine-learning method, better than the threshold method of 97.54%, which is consistent with the experimental result. This work indicates that the machine-learning method can be a strong potential candidate for alleviating the restrictions of stably achieving high-fidelity and high-accuracy singlet-triplet-state readout in large-scale quantum computing.

arXiv:2309.09723•Physical Review Applied•Mar 13, 2024•8 citations

Transmon qubit readout fidelity at the threshold for quantum error correction without a quantum-limited amplifier

Liangyu Chen, Hang-Xi Li, Yong Lu +17 more

Abstract High-fidelity and rapid readout of a qubit state is key to quantum computing and communication, and it is a prerequisite for quantum error correction. We present a readout scheme for superconducting qubits that combines two microwave techniques: applying a shelving technique to the qubit that reduces the contribution of decay error during readout, and a two-tone excitation of the readout resonator to distinguish among qubit populations in higher energy levels. Using a machine-learning algorithm to post-process the two-tone measurement results further improves the qubit-state assignment fidelity. We perform single-shot frequency-multiplexed qubit readout, with a 140 ns readout time, and demonstrate 99.5% assignment fidelity for two-state readout and 96.9% for three-state readout–without using a quantum-limited amplifier.

arXiv:2208.05879•npj Quantum Information•Mar 16, 2023•103 citations
Abstract Single-shot readout is a key component for scalable quantum information processing. However, many solid-state qubits with favorable properties lack the single-shot readout capability. One solution is to use the repetitive quantum-non-demolition readout technique, where the qubit is correlated with an ancilla, which is subsequently read out. The readout fidelity is therefore limited by the back-action on the qubit from the measurement. Traditionally, a threshold method is taken, where only the total photon count is used to discriminate qubit state, discarding all the information of the back-action hidden in the time trace of repetitive readout measurement. Here we show by using machine learning (ML), one obtains higher readout fidelity by taking advantage of the time trace data. ML is able to identify when back-action happened, and correctly read out the original state. Since the information is already recorded (but usually discarded), this improvement in fidelity does not consume additional experimental time, and could be directly applied to preparation-by-measurement and quantum metrology applications involving repetitive readout.
Machine learning is a powerful tool in finding hidden data patterns for quantum information processing. Here, we introduce this method into the optical readout of electron-spin states in diamond via single-photon collection and demonstrate improved readout precision at room temperature. The traditional method of summing photon counts in a time gate loses all the timing information crudely. We find that changing the gate width can only optimize the contrast or the state variance, not both. In comparison, machine learning adaptively learns from time-resolved fluorescence data and offers the optimal data processing model that elaborately weights each time bin to maximize the extracted information. It is shown that our method can repair the processing result from imperfect data, reducing 7% in spin readout error while optimizing the contrast. Note that these improvements only involve recording photon time traces and consume no additional experimental time, and they are, thus, robust and free. Our machine learning method implies a wide range of applications in the precision measurement and optical detection of states.
While single-shot detection of silicon spin qubits is now a laboratory routine, the need for quantum error correction in a large-scale quantum computing device demands a quantum non-demolition (QND) implementation. Unlike conventional counterparts, the QND spin readout imposes minimal disturbance to the probed spin polarization and can therefore be repeated to extinguish measurement errors. Here, we show that an electron spin qubit in silicon can be measured in a highly non-demolition manner by probing another electron spin in a neighboring dot Ising-coupled to the qubit spin. The high non-demolition fidelity (99% on average) enables over 20 readout repetitions of a single spin state, yielding an overall average measurement fidelity of up to 95% within 1.2 ms. We further demonstrate that our repetitive QND readout protocol can realize heralded high-fidelity (>99.6%) ground-state preparation. Our QND-based measurement and preparation, mediated by a second qubit of the same kind, will allow for a wide class of quantum information protocols with electron spins in silicon without compromising the architectural homogeneity.
Accurate quantum state readout is crucial for error correction and algorithms, but measurement errors are detrimental. Readout fidelity is typically limited by a poor signal-to-noise ratio (SNR) and energy relaxation (T1T_1T1​ decay), a significant problem for superconducting qubits. While most approaches classify results using clustering algorithms on integrated readout signals, these methods cannot distinguish a qubit that was initially in the ground state from one that decayed to it during measurement. We instead propose using machine learning (ML) on the raw, non-integrated analog signal. We apply time-series classification models, such as a long short-term memory (LSTM) network, to the full data trajectory. We find that our LSTM model, combined with filtering and feature engineering, consistently outperforms clustering. The largest improvements come from reclassifying points in the boundary regions between clusters. These points correspond to atypical measurement records, likely due to transient or noisy features lost during data integration. By retaining temporal information, sequence-aware models like LSTMs can better discriminate these trajectories, whereas clustering methods based on integrated values are more prone to misclassification.
Quantum systems, including superconducting circuits, trapped ions, quantum dots, solid-state defects, etc., have achieved considerable advancements in readout fidelity. However, the widely used threshold method disregards the importance of temporal characteristics of the signal during continuous measurements, leading to information loss. Here, we applied Fisher information to quantify the potential enhancement by using the temporal information. We proposed an optimal data processing method and derived the condition for equality to the error constrained by Fisher information. Applying this method to the single-shot readout of a [Formula: see text] nuclear spin of nitrogen-vacancy center marks a 33.8 ± 1.2% error reduction compared to the threshold method, improving the fidelity to 99.649(5)%. The versatility of our method is further demonstrated through the single-shot readout of charge states. This work illustrates the advantage of using photon arrival time for improving readout fidelity without upgrading experimental hardware. The method of using Fisher information to analyze readout fidelity also holds promise for broad applicability to other systems.
High-fidelity singlet-triplet state readout is essential for large-scale quantum computing. However, the widely used threshold method of comparing a mean value with the fixed threshold will limit the judgment accuracy, especially for the relaxed triplet state, under the restriction of relaxation time and signal-to-noise ratio. Here, we achieve an enhanced latching readout based on Pauli spin blockade in a Si-MOS double quantum dot device and demonstrate an average singlet-triplet-state readout fidelity of 97.59% by the threshold method. We reveal the inherent deficiency of the threshold method for the relaxed triplet-state classification and introduce machine learning as a noise-resilient and relaxation-independent readout method to reduce the misjudgment. The readout fidelity for classifying the simulated single-shot traces can be improved to 99.67% by the machine-learning method, better than the threshold method of 97.54%, which is consistent with the experimental result. This work indicates that the machine-learning method can be a strong potential candidate for alleviating the restrictions of stably achieving high-fidelity and high-accuracy singlet-triplet-state readout in large-scale quantum computing.
Abstract High-fidelity and rapid readout of a qubit state is key to quantum computing and communication, and it is a prerequisite for quantum error correction. We present a readout scheme for superconducting qubits that combines two microwave techniques: applying a shelving technique to the qubit that reduces the contribution of decay error during readout, and a two-tone excitation of the readout resonator to distinguish among qubit populations in higher energy levels. Using a machine-learning algorithm to post-process the two-tone measurement results further improves the qubit-state assignment fidelity. We perform single-shot frequency-multiplexed qubit readout, with a 140 ns readout time, and demonstrate 99.5% assignment fidelity for two-state readout and 96.9% for three-state readout–without using a quantum-limited amplifier.