Systems and methods for calibrating qudit read-out

A method for calibrating qudit read-out signals by cycling through computational states without initialization optimizes parameters based on measurement results, addressing the inefficiencies of existing qubit read-out calibration methods and achieving faster, efficient calibration.

WO2026235545A1PCT designated stage Publication Date: 2026-11-12ORANGE QUANTUM SYSTEMS HOLDING BV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/NL2026/050123
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-07
Filing Date
2026-05-04
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing qubit read-out calibration methods are time-consuming due to the need for extensive initialization and are not suitable for fast characterization and pulse calibration, especially in restless tune-up protocols where initialization steps are skipped.

Method used

A method for calibrating qudit read-out signals by cycling the qudit through computational basis states without intermediate initialization, using a sequence of control and read-out signals to generate response signals, and optimizing parameters based on measurement results to minimize measurement-induced transitions and distinguish states.

Benefits of technology

The method significantly reduces calibration time by eliminating intermediate initialization steps, allowing for a ten-fold increase in measurements in less time and achieving faster, efficient calibration of qudit read-out signals.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure NL2026050123_12112026_PF_FP_ABST
    Figure NL2026050123_12112026_PF_FP_ABST
Patent Text Reader

Abstract

Methods and systems are disclosed for calibrating a qudit read-out signal. The method comprises configuring a control signal which cycles the qudit through k computational basis states, and configuring a read-out signal based on optimisation parameters. The method further comprises obtaining measurement results by applying control signals comprising repetitions of the control signal and the read-out signal, and measuring the response signals, without initialising the qudit in between. The measurement results are substantially equally assigned to k measurement bins based on the respective values of the measurement results. Provisional labels are assigned to each measurement result to obtain provisional labels, based on the measurement bin to which a given measurement result has been assigned. The method further comprises computing an error based on a deviation between the provisional labels and expected labels, based on the control signals, and updating the optimization parameters based on the error.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] WO41704-Sv / TD

[0002] Systems and methods for calibrating qudit read-out

[0003] Technical field

[0004] The disclosure relates to calibrating qubit control, and in particular, though not exclusively, to methods and systems for calibrating a read-out operation of a qubit, and to a computer program product enabling a quantum device to perform such methods.

[0005] Background

[0006] The benchmarking and operation of quantum devices such as quantum processing units (QPUs), requires extensive, time-consuming calibration protocols to ‘tune up’ control operations. The calibration protocols typically require a well-defined initial state of the system, which is conventionally achieved by either qubit relaxation or active reset. As advancements are made in QPU quality, qubit initialization by relaxation, which typically takes five times the qubit relaxation time, will provide a major contribution to the total experiment duration, making it not suitable for fast characterization and pulse calibration. Although active reset delivers rapid and reliable qubit initialization, it requires previously calibrated operations and control hardware capable of fast feedback.

[0007] Superconducting qubits are generally measured by a process called dispersive readout. The qubit is dispersively coupled to a read-out resonator whose resonant frequency depends on the state of the qubit. In order to read out the state of the qubit, a microwave pulse is applied to the resonator and the returned signal is measured via heterodyne spectroscopy. Typically, the measured signal (represented as a complex value) in the l / Q plane will follow a different statistical distribution depending on the state of the qubit. In order to discriminate between different qubit states, this heterodyne measurement must be calibrated so that different qubit states can be distinguished based on the measured signal in the l / Q plane. The purpose of the calibration process is to fine-tune the parameters of the read-out spectroscopy pulse such that the contrast between the distributions belonging to different qubit states is maximised, allowing for high-fidelity read-out of the qubit state.

[0008] Read-out calibration is typically done by preparing the qubit successively in the |0) and the |1) state (higher states of the qubit can also be included if desired). The dispersive measurement is carried out many times in both of these states, allowing to fit a statistical distribution to the data in both states (typically these distributions are assumed to take a Gaussian form). The two-dimensional l / Q data can then be rotated so that the means of boththese distributions lie on the x-axis and a threshold is set such the qubit state can be determined depending on which side of the threshold the measured l / Q value lies. The optimal threshold will be the point at which the total probability that a measured l / Q value falls on the wrong side of the threshold is minimised. When calibrating the dispersive readout process, the parameters of the read-out microwave pulse can be optimised so as to minimise this error probability.

[0009] Although this method works well in enabling high fidelity dispersive read-out of superconducting qubits, the calibration process is slow and time-consuming, due to the need to initialise the qubit in both the |0) and |1) state many times in order to measure the statistical distributions for each state. This provides a significant overhead when tuning up a quantum chip. Passive qubit initialisation protocols take several milliseconds to complete, while with active qubit reset this can be reduce to tens or hundreds of microseconds. With many thousands of iterations required to measure the readout fidelity, this can still present a significant overhead in qubit calibration.

[0010] For general quantum chip tune-up protocols, Orange Quantum Systems has developed a system of restless tune-up, as described in WO 2024 / 231488 A1 , which removes the need for qubit initialisation on each step of the process. In this method, instead of initialising a qubit in a given state before implementing a quantum gate and checking that the final state is correct, we simply implement gates back to back with quantum state measurements and check that the gate had the correct effect on the previously measured quantum state.

[0011] However, this process does not work for readout calibration because in order to measure the readout fidelity, one needs to be able to fit two different Gaussian distributions to the measured l / Q values corresponding to the |0) and |1) states. This requires a-priori knowledge of which data points corresponds to which state. In previous implementations, this information was available because the qubit was initialised in either |0) or 11 > before every run of the experiment, but in restless tune-up, the initialisation step is skipped, so the two Gaussians cannot be distinguished a-priori.

[0012] Hazra et al., ‘Benchmarking the Readout of a Superconducting Qubit for Repeated Measurements’, Physical Review Letters 134 (2025), 100601 describe readout-induced leakage benchmarking for superconducting qubits.

[0013] Hence, there is a need in the art for a fast method to calibrate a qubit read-out operation.

[0014] Summary

[0015] It is an aim of the embodiments in this disclosure to provide a system and method for calibrating a qudit that avoids, or at least reduces the drawbacks of the prior art.In a first aspect, embodiments in this disclosure relate to a method for calibrating a qudit, in particular for calibrating a qudit read-out signal. The method comprises configuring a qudit control signal, the qudit control signal being configured to cycle the qudit through k computational basis states associated with the qudit, k being an integer greater than 1 , and configuring a qudit read-out signal based on one or more first optimization parameters.

[0016] Application of the qudit read-out signal to the qudit generates a response signal indicative of a computational basis state of the qudit. The method further comprises obtaining a sequence of measurement results by applying a sequence of control signals, the sequence of control signals comprising a repetition of the qudit control signal and the qudit read-out signal, and measuring the response signals, generally without initialising the qudit between measuring the response signal and applying the qudit control signal. The measurement results are assigned to k measurement bins based on the respective values of the measurement results, such that each of the k measurement bins comprises a substantially equal number of measurement results. Provisional labels are assigned to each of the measurement results in the sequence of measurement results to obtain a sequence of provisional labels, the provisional labels being based on the measurement bin to which a given measurement result has been assigned. The method further comprises computing an error based on a deviation between the sequence of provisional labels and an expected sequence of labels, the expected sequence of labels being based on the sequence of control signals, and updating the one or more first optimization parameters based on the error.

[0017] Thus, the output of each action is used as input for the subsequent action; i.e., the state of the qudit after application of a control signal (e.g., a ‘qudit control signal’) from the sequence of control signals, is also the state of the qudit upon application of a subsequent control signal (e.g., a ‘read-out signal’) from the sequence of control signals. This is also known as a ‘restless’ procedure. Consequently, the method is particularly suitable to optimise projective quantum measurements that exhibit quantum non-demolitionness (QND), i.e., measurements that do not, or at least should not, cause a measurement-induced transition of the qudit.

[0018] By not using intermediated initialisation steps, the methods described herein are much faster than known methods wherein the qudit is initialised in a known state before a measurement signal is being applied. Moreover, contrary to such known methods, the methods described herein are also sensitive to measurement-induced transitions.

[0019] A deviation between the sequence of provisional labels and an expected sequence of labels can be determined by comparing pairs or short sequences of labels, e.g. by determining a difference between two subsequent measurement results. It is not always necessary to explicitly generate the expected sequence of labels; for example, the expecteddifference between two subsequent labels may always be 1 (modulo a number of states), and hence any value different from 1 may indicate an error.

[0020] In principle, the method as described above only optimises the qudit read-out signal, e.g., to obtain maximally distinguishable outputs for different computational basis states, to minimise measurement-induced transitions, ora combination thereof. However, the method does not necessarily provide information which provisional label should be associated with which computational basis state.

[0021] Hence, in some embodiments, the method further comprises, after optimising the qudit read-out signal, initialising the qudit in a known computational basis state, applying the qudit read-out signal, and measuring the response signal; and assigning a definitive label to each measurement bin based on the measured response signal and the known computational basis state.

[0022] Typically, the sequence of control signals consists of (only) qudit control signals for causing the qudit to transition between basis states, and qudit read-out signals. The qudit control signal is generally the same in each repetition. As used herein, a single control signal may comprise multiple pulses; for example, for a qutrit, the qutrit control signal could consist of a pulse calibrated for a transition between a |0) state and a |1) state (and vice versa) followed by a pulse calibrated for a transition between a |0) state and a |2) state (and vice versa). This same pair of pulses may then be applied repeatedly, alternated with measurement operations. A measurement operation may comprise application of a pulse to the qudit and an acquisition of a response from the qudit.

[0023] It can generally be assumed that the qudit control signal is already calibrated, or at least that the error in the qudit control signal is substantially smaller than in the measurement signal. In some embodiments, a calibration process may comprise alternating phases of qudit control signal calibration and qudit read-out signal calibration.

[0024] In general, the number of measurement results in a sequence of measurement results should be sufficient to allow for reliable statistical analysis. It is noted that because the qudit is not initialised in between measurements, more measurements may be performed in less time, than in methods where the qudit is initialised in a known state in between consecutive measurements (typically, cycles consisting of initialisation - operation(s) - measurement). Increasing the amount of measurements (without increasing the total calibration time) also allows detection of errors that occur less often.

[0025] In an embodiment, the expected sequence of labels is symmetric with respect to permutation of the labels. As noted above, in general, it is not known to which computational basis state each provisional label corresponds. By ensuring that the expected sequence of labels is symmetric with respect to permutation of the labels, the analysis is insensitive to the assignment of labels.In an embodiment, assigning the measurement results to the k measurement bins comprises performing a principal-component analysis, PCA, on the measurement results and dividing the measurement results into bins along a direction of a principal component. An advantage of using principal component analysis, in particular for qubits, is that it is very fast and lightweight. Other embodiments, in particular embodiments with qudits having more than two computational basis states, may use different classification algorithms, such as a support vector machine or other known two-dimensional classification algorithms, or they can rely on (frequency) multiplexed measurements to obtain a higher-dimensional dataset on which a higher-dimensional classification algorithm may be performed, e.g., a higher-order principalcomponent analysis.

[0026] In an embodiment, obtaining measurement results comprises alternating a single qudit control signal and a single qudit read-out signal. This results in a sequence that is straightforward to analyse. Other embodiments may use different sequences, e.g., two consecutive read-out signals in between each qudit control signal. In general, the control sequence may be optimised based on the expected errors.

[0027] In an embodiment, computing the error comprises computing a cost function configured to penalize measurement results if the sequence of provisional labels deviates from the expected sequence of labels. For example, the cost function may be given by

[0028] IW-1

[0029] eR=TT

[0030] 1V — — 71 £ / — I -mn-i)m°d k)

[0031]

[0032] n=l

[0033] wherein eRrepresents an error, N represents a number of measurements, mnrepresents a binned measurement result (typically represented by a natural number used as an index) corresponding to measurement number n, and f {0, ... , k} K is a weight function. Again, k is the number of basis states of the qudit. For qudits with more than two computational basis states, the error may be computed for all permutations of provisional labels, and the smallest value may be selected as the computed error.

[0034] In the case of qubits (k = 2), the error function may be given by

[0035] IW-1

[0036] eR=I MV - — 71 £ / — I =mn-l)

[0037]

[0038] n=l

[0039] wherein eRrepresents an error, N represents a number of measurements, mnrepresents a binned measurement result corresponding to measurement number n, and

[0040] f {TRUE, FALSE} -> K is a weight function.In an embodiment, the method further comprises repeating the obtaining of the measurement results by repeatedly applying the qudit control signal and the qudit read-out signal and measuring the response signal, the binning of the measurement results, the computing of the error, and the updating of the one or more first optimization parameters until the error meets one or more optimization conditions.

[0041] The updating of the one or more first optimization parameters may be based on an optimisation algorithm.

[0042] In an embodiment, the qudit control signal is based on one or more second optimization parameters, the method further comprising updating the one or more second optimization parameters based on the error.

[0043] In an embodiment, the qudit is a qubit, and the qudit control signal is a qubit flip signal configured to flip the qubit from a first computational basis state, e.g. a |0) state, to a second computational basis state different from the first computational basis state, e.g. a |1) state. Most current quantum processors are limited to manipulating qubits.

[0044] A further aspect of this disclosure relates to a hybrid computing system comprising a classical computer and a quantum processor. The quantum processor comprises a plurality of qudits, e.g., qubits, at least one of which is to be calibrated. The qudits may be superconducting qudits, semiconductor-based spin qudits, or any other qudit modality that has a projective and (approximately) non-destructive read-out mechanism.

[0045] The classical computer comprises at least one processor and a memory storing computer-readable instructions, the memory being communicatively coupled to the at least one processor. The classical computer may be configured to control (directly or indirectly, e.g., via intermediate hardware as is known in the art) the quantum processor.

[0046] The at least one processor is configured to, in response to executing the computer-readable instructions: configuring a qudit control signal, the qudit control signal being configured to cycle the qudit through k computational basis states associated with the qudit, k being an integer greater than 1 ; configuring a qudit read-out signal based on one or more first optimization parameters, application of the qudit read-out signal to the qudit generating a response signal indicative of a computational basis state of the qudit; obtaining, using the quantum processor, a sequence of measurement results by applying a sequence of control signals, the sequence of control signals comprising a repetition of the qudit control signal and the qudit read-out signal, and measuring the response signals, without initialising the qudit between measuring the response signal and applying the qudit control signal; assigning the measurement results to k measurement bins based on the respective values of the measurement results, each of the k measurement bins comprising a substantially equal number of measurement results; assigning provisional labels to each of the measurement results in the sequence of measurement results to obtain a sequence of provisional labels,the provisional labels being based on the measurement bin to which a given measurement result has been assigned; computing an error based on a deviation between the sequence of provisional labels and an expected sequence of labels, the expected sequence of labels being based on the sequence of control signals; and updating the one or more first optimization parameters based on the error.

[0047] The step of obtaining the sequence of measurement results may comprise configuring the quantum processor to apply the sequence of control signals, and receiving the measurement results from the quantum processor.

[0048] In further embodiments, the system is further configured for executing any one or more of the method steps described above.

[0049] One aspect of this disclosure relates to a classical computer system or hybrid quantum computer system comprising a computer readable storage medium having computer readable program code embodied therewith, and a processor, preferably a microprocessor, coupled to the computer readable storage medium, wherein responsive to executing the computer readable program code, the processor is configured to perform any of the methods described herein.

[0050] One aspect of this disclosure relates to a computer program or suite of computer programs comprising at least one software code portion or a computer program product storing at least one software code portion, the software code portion, when run on a classical computer system or hybrid quantum computer system, being configured for executing any of the methods described herein.

[0051] One aspect of this disclosure relates to a non-transitory computer-readable storage medium storing at least one software code portion, the software code portion, when executed or processed by a computer, configures the computer or hybrid computing system to perform any of the methods described herein.

[0052] As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or computer program product. Accordingly, aspects of the present embodiments may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system”. Functions described in this disclosure may be implemented as an algorithm executed by a microprocessor of a computer. Furthermore, aspects of the present embodiments may take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied, e.g., stored, thereon.

[0053] Any combination of one or more computer readable medium(s) may be utilized. The computer readable medium may be a computer readable signal medium or a computerreadable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non- exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fibre, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0054] A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0055] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fibre, cable, RF, etc., or any suitable combination of the foregoing. Computer program code for carrying out operations for aspects of the present embodiments may be written in any combination of one or more programming languages, including a functional or an object oriented programming language such as Java, Scala, C++, Python or the like and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may execute entirely on the user’s computer, partly on the user’s computer, as a stand-alone software package, partly on the user’s computer and partly on a remote computer, or entirely on the remote computer, server or virtualized server. In the latter scenario, the remote computer may be connected to the user’s computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0056] Aspects of the present embodiments are described below with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the embodiments. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in theflowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0057] These computer program instructions may be provided to a processor, in particular a microprocessor or central processing unit (CPU), or graphics processing unit (GPU), of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer, other programmable data processing apparatus, or other devices create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0058] These computer program instructions may also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function / act specified in the flowchart and / or block diagram block or blocks.

[0059] The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0060] The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present embodiments. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations of blocks in the block diagrams and / or flowchart illustrations, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

[0061] The embodiments will be further illustrated with reference to the attached drawings, which schematically will show embodiments according to the invention. It will be understood that the invention is not in any way restricted to these specific embodiments. Identical reference signs refer to identical, or at least similar elements.Brief of the figures

[0062] Fig. 1 A is a schematic illustration of a read-out calibration method according to a known method and Fig. 1B is a schematic illustration of a read-out calibration method according to an embodiment;

[0063] Fig. 2A is a histogram of simulated measurement output of a read-out calibration method according to a known method, Fig.2B and 2C are histograms of a simulated measurement output of a read-out calibration method according to an embodiment, and Fig.

[0064] 2D and 2E show measurement results of a read-out calibration method according to an embodiment;

[0065] Fig. 3 illustrates data analysis of measurement results using a method according to an embodiment;

[0066] Fig. 4 shows a run-time comparison of a read-out calibration method according to a known method with a read-out calibration method according to an embodiment;

[0067] Fig. 5A is a schematic illustration of a read-out calibration method according to an embodiment, and Fig. 5B and 5C schematically illustrate measurement output of various stages of the method;

[0068] Fig. 6 is a flow chart of a method according to an embodiment; and

[0069] Fig. 7A and 7B are block diagrams of a system according to an embodiment.

[0070] Detailed description

[0071] Fig. 1 A is a schematic illustration of a read-out calibration method according to a known method. The known method comprises two different subroutines 100, 110 for different computational basis states. A first subroutine initialises 102 a qubit in the zero-state, and then performs a measurement 106. The measurement may cause an (unwanted) change in the state of the qubit. Therefore, in the next round, the qubit is again initialised 102 before being measured again 106. The result is a collection of measurements 108, typically in the / Q-plane, which are all associated with the zero-state. Here, I denotes the in-phase data and Q denotes the quadrature data.

[0072] A second subroutine 110 similarly initialises 112 a qubit in the zero-state, then flips 114 the qubit to the one-state (in this example by using an X-gate), and subsequently performs a measurement 116. The measurement may again cause an (unwanted) change in the state of the qubit. Therefore, in the next round, the qubit is again initialised 112 before being flipped114 and measured again 116. The result is a collection of measurements 118, in the same space as for the first subroutine, which are all associated with the one-state.

[0073] The measurements for the zero-state and the one-state may be analysed separately, and can be used to optimise the read-out signal 106,116 in an iterative way.

[0074] However, this method has several drawbacks. First of all, the initialisation steps 102,112 take a lot of time, typically over 80% or 90% of the entire cycle 100,110. However, leaving out the initialisation would lead to a series of erroneous measurements, as a single error propagates until an ‘opposite’ error occurs. Moreover, this method is insensitive to measurement induced state changes, which can be relevant for protocols relying on nondestructive (intermediate) measurements.

[0075] Fig. 1B is a schematic illustration of a read-out calibration method according to an embodiment. This method comprises only a single routine 120 for all basis states. The qubit may optionally be initialised 122 prior to the first iteration. Subsequently, the qubit is alternately subjected to a bit-flip operation 124,134 and a measurement operation 126,136, without intermediate initialisation. That is, the output state 127 of the measurement operation 126 is the input state of the subsequent bit-flip operation 134. The result is a single collection of measurement results 128, of which the corresponding ‘real’ state is unknown. The processing of this collection of measurement results is described in more detail below with reference to Fig. 2 and 3.

[0076] As an example, in a situation without any errors, the initialisation 122 may result in an initial state 123 m0= |0). The bit-flip operation 124 causes the qubit to flip to the |l)-state (not shown). The measurement operation 126 does not change the state of the qubit, so the output state of the measurement operation 127 equals mj = |1). This is also the input state of the next bit-flip operation 134, which causes the qubit to flip to the |0)-state (not shown). Again, the measurement operation 136 does not change the state of the qubit, and hence, the output of the second measurement operation is m2= |0). Thus, in the ideal case, mn= |0) for n even and mn= |1) for n odd, i.e. , the expected measurement results take the form 101010101010... (starting at m1, as m0is not measured in de depicted example). However, in practice, errors occur, and hence it is not possible to determine the state based on the parity of the measurement index; for example, the actual measured states could be 101001010101....

[0077] It is noted that in the current example, two iterations are shown, to facilitate the description of the process; the process could also have shown, more succinctly, as only comprising blocks 123,124,126.

[0078] It is noted that although a sequence of alternating bit-flips and measurements may be easiest to implement, different sequences may also be used. E.g., after each bit-flip, two measurements may be performed, leading to an (intended) sequence 110011001100.... Ingeneral, however, the intended sequence is symmetric with respect to permutation of the states; that is, by changing the labels assigned to the states, the same sequence is obtained up to a permutation of the states.

[0079] Fig. 2A is a histogram of the measurement results of the two known subroutines shown in Fig. 1A. The bars 208 with the downward hatching correspond to the measurements 108 of the qubit in the zero state, and the bars 218 with the upward hatching correspond to the measurements 118 of the qubit in the one state. Although two distinct peaks can be identified, there is also substantial overlap between the two histograms. A typical optimisation process would aim to minimise the overlap.

[0080] Fig. 2B is a histogram of the measurement results of the read-out calibration method according to the embodiment shown in Fig. 1B. Here, a single broad distribution 228 is obtained, without clearly identifiable peaks; and it is a priori unknown which measurement results from which (expected) computational basis state.

[0081] Thus, as is shown in Fig. 2C, once the collection of measurements has been obtained, provisional labels may be assigned to each measurement. In principle, any known classification algorithm may be used, depending on the dimensionality of the measurement results and the number of computational basis states. For example, if only a single parameter is measured (one dimension), the measurement results may be divided into two groups based on, e.g., the median, mean, or another similar statistical representative of the measurement results. Although more elaborate algorithms may also be used (e.g., based on fitting parametrised functions, e.g., Gaussian curves), lightweight methods like those based on the median or mean are typically faster and less computationally expensive, while yielding equal or at least similar results, due to the fact that both measurement outcomes should occur in the same number (assuming an errorless system). In the depicted example, the measurements are based in two equal groups 238,248 based on the l-value (horizontal axis in plot 128), using the median value as the divider or threshold.

[0082] In higher-dimensional systems, such as the commonly used IQ-plane, known dimension-reducing methods may be used, such as principal component analysis. The results may then be divided along the axis of the principal component, as described for the one-dimensional case.

[0083] Thus, the measurement results may be divided into a first measurement bin (e.g., smaller than the median or mean), and a second measurement bin (e.g., larger than the median or mean). As used herein, bins refer to mutually disjunct sets of elements. Each measurement bin is assigned a provisional label, e.g., ‘high’ and ‘low’. At this point, it is generally not known which label corresponds (best) to which computational basis state.

[0084] Applying these provisional labels to the sequence of measurement results, a sequence of provisional labels is obtained. This sequence of provisional labels is then compared to asequence of expected outcomes. As the relation between the provisional labels and the computational basis states is not known, this is generally done by comparing pairs of subsequent measurement results. For example, in an embodiment where a sequence of alternating states is expected, an error may be determined if two subsequent provisional labels are identical to each other; and in an embodiment where a sequence of alternating equal pairs is expected (i.e. 00110011...), an error may be determined when a provisional label is either equal to both neighbouring labels or different from both neighbouring labels.

[0085] Fig. 2D and 2E show (real, i.e., not simulated) measurement results of a read-out calibration method according to an embodiment. In particular, Fig. 2D shows a conventional analysis method, wherein (double) Gaussian distributions are fitted to the |0) and |1) datasets respectively. However, this is only possible with a priori knowledge of which datapoints correspond to which qubit state (obtained from the initialisation in each step). Fig.

[0086] 2D shows Gaussian curves fitted to l / Q distributions which have been rotated onto the x-axis, along with a threshold at 1.34 mV. This fitting of (double) Gaussians is relatively computationally expensive, and time-consuming.

[0087] In the methods described herein, there is no need to fit such Gaussians (or other functions). Instead, by performing Principal Component Analysis (PCA) to separate the measurement data, the need for a priori knowledge of the qubit state for each measurement is removed. PCA allows to find the vectors of maximal and minimal deviation for the collection of measurement data, here in / Q-space. For a general distribution of |0) and |1) data, the vector of maximal variance (corresponding to the eigenvector of the covariance matrix with the largest eigenvalue) will correspond approximately to the displacement of the centre of one quantum state distribution to the other. Therefore, by projecting onto this vector, a one-dimensional distribution is obtained with two main peaks corresponding to the computational basis states (although it is not known at this stage which is which). A threshold can be set that (approximately or exactly) half of the datapoints are on one side and half on the other. Then, a cost function can be computed for this dataset (as described in more detail below) and the read-out pulse parameters can be adjusted (varied) in order to optimise (typically minimise) this cost function. Removing the need for qubit initialisation on every step leads to a significant speed-up in this optimisation process.

[0088] Fig. 2E shows two typical l / Q diagrams that resulting from a read-out calibration process. Panel 260 shows a group of datapoints corresponding to qubits in the |0) basis state, while panel 270 shows a group of datapoints corresponding to qubits in the |1) basis state. A threshold 266 can be set in between these two distributions in order to discriminate between the two states, as shown by the solid line (the same threshold line is shown in both panels). Errors can come about because of the overlap between these two distributions, resulting in datapoints, that fall on the ‘wrong’ side of the threshold. In the current example,results 262 below the threshold 266 would be (correctly) assigned to the |0) state, while results 264 above the threshold 266 would be (wrongly) assigned to the |1) state; and similarly, results 272 below the threshold 266 would be (wrongly) assigned to the |0) state, while results 274 above the threshold 266 would be (correctly) assigned to the |1) state. This error can be minimised by optimising the readout pulse parameters. Some additional readout operations may be performed on qubits that have been initialised in a fiducial state (typically a computational basis state) in order to associate the measurement bins with computational basis states of the qubit.

[0089] In the methods described herein, a similar distinction between two sets of measurement results is made. However, instead of determining the threshold by fitting double-Gaussians to distributions of known basis states, the threshold is determined based on a classification algorithm, e.g., a principal-component analysis. This may result in a (slightly) different threshold value. Additionally, provisional labels are assigned to each side of the threshold, e.g., ‘A’ and ‘B’, where it is not (yet) known whether ‘A’ corresponds to |0) and ‘B’ corresponds to 11), or vice versa.

[0090] Fig. 3 illustrates data analysis of measurement results using a method according to an embodiment. Restless single-shot readout calibration raw measurement analysis. Top row shows raw measurement data, and the bottom row shows processed data using a principal component analysis method, which essentially rotates the data along the axis of most signal. A threshold is determined through the centre of the data set, and results on opposite sides of the threshold are assigned different provisional labels.

[0091] Cost function values are computed by comparing the sequence of provisional labels with an expected sequence of labels. In this case, a series of alternating labels is expected, as control sequence 120 as shown in Fig. 1B was used to obtain the measurement results. The error was computed using

[0092] IW-1

[0093] eR=I MV — — 71 £ / — I=

[0094]

[0095] n=l

[0096] wherein eRrepresents the error, N represents a number of measurements, mnrepresents a binned measurement result corresponding to measurement number n, and

[0097] f {TRUE, FALSE} -> K is a weight function.

[0098] The cost function magnitudes are computed after thresholding the rotated data shown in the bottom row, and equal 0.54, 0.23, and 0.55 for the -6 dBm (left column), -30 dBm (middle column), and -52 dBm (right column) read-out pulse power regimes, respectively.Fig. 4 shows a run-time comparison per iteration between the conventional and restless calibration methods. Here an iteration is considered as the execution of the test protocol (measurement schedule), data analysis, and setting of the new read-out parameters for the subsequent iteration. The restless single-shot read-out calibration method (i.e., using a method as described herein) demonstrates a threefold decrease in run time per iteration compared to the conventional approach.

[0099] In this example, one schedule of the conventional read-out calibration method was constructed of 2 measurement operations, with 1024 shots per measurement. In contrast, the restless read-out calibration method’s schedule was constructed of 20 measurement operations, with 1000 shots per measurement. Despite the significantly larger number of measurement operations in the restless schedule, the duration of the experiment remains shorter. Thus, due to the elimination of the initialisation in each step, a roughly ten-fold increase in measurements was achieved in less time, compared to the conventional method. The number of required measurement operations in the restless schedule may be further optimised to achieve a reasonable balance between execution time and signal-to-noise ratio. Potentially, this could lead to an additional speedup if fewer measurement operations are found to be sufficient.

[0100] The primary speedup per iteration is achieved in the data analysis run time and a reduction in the additional overhead. The principal component analysis method is orders of magnitude faster compared to the double Gaussian peak fitting method used for the data analysis by the conventional read-out calibration protocol.

[0101] Fig. 5A is a schematic illustration of a read-out calibration method according to an embodiment. In this example, the qudit is a qutrit with three computational basis states (i.e., k = 3), denoted 10), 11), and |2). After an optional initialisation 502, a composite qudit control signal 504,506 is alternated with a measurement signal 508. This sequence is configured to cycle the qudit through computational basis states 503.

[0102] In this specific example, the qudit control signal consists of two pulses, a first pulse 504 tuned to a transition between the |l)-state and the |2)-state followed by a second pulse 506 tuned to a transition between the |0)-state and the |l)-state. If the qutrit is in the |0)-state, the first pulse will have no effect, and the second pulse will cause the qutrit to transition from the |0)-state to the |l)-state; if the qutrit is in the |l)-state, the first pulse will cause the qutrit to transition from the |l)-state to the |2)-state, and the second pulse will have no effect; and if the qutrit is in the |2)-state, the first pulse will cause the qutrit to transition from the |2)-state to the |l)-state, and the second pulse will cause the qutrit to transition from the |l)-state to the |0)-state. Thus, the combination of the first and second pulse, will cause the qutrit to transition from the |0)-state to the |l)-state, from the |l)-state to the |2)-state, and from the |2)-state to the |0)-state (via the |l)-state). Thus, absent any errors, repeated application ofthe qudit control signal will cause the qutrit to cycle through all three computational basis states, like 012012012... In other embodiments, the qudit control signal may comprise more, fewer, or different pulses, or a different sequence of the same pulses. In general for qudit systems with k basis states, a sequence of k - 1 pulses tuned to the transitions | Zc — 1> \k}, ..., |0) |1) may be used.

[0103] Before calibration, the measurement results may look as in Fig. 5B. The measurement results may be divided into k = 3 substantially equally sized bins, and provisional labels may be assigned to each measurement result based on the bin to which the measurement result has been assigned, analogous to the case for k = 2 described above. Known classification algorithms may be used to determine the different bins, e.g., a higher-dimensional PCA method.

[0104] Fig. 5C is a schematic illustration of the measurement results after calibration. In this example, two thresholds 532,534 are determined, one separating the |0) state from the |1) and 12) states, and the other separating the |2) state from the |0) and |1) states. Other implementations may lead to different divisions.

[0105] Fig. 6 is a flow chart of a method according to an embodiment. The method may be executed by a classical computer configured to control a quantum device that is being calibrated.

[0106] A step 602 comprises configuring a qudit control signal that is configured to cycle the qudit through d computational basis states associated with the qudit. Here, d being an integer greater than 1 , e.g., k = 2 for a qubit, k = 3 for a qutrit, and so on. For example, for k = 2, the qudit control signal can be a qubit flip signal, configured to flip the qudit between to computational basis states, as is known in the art. For k > 2, the qudit control signal may comprise several pulses, as described above with reference to Fig. 5A.

[0107] In some embodiments, the qudit control signal is maintained throughout the read-out calibration procedure. In other embodiments, the qudit control signal is based on one or more qudit-control optimization parameters, which may be optimised along with the one or more read-out optimization parameters (described in the next step).

[0108] A step 604 comprises configuring a qudit read-out signal based on one or more readout optimization parameters. The application of the qudit read-out signal to the qudit generates (results in) a response signal indicative of a computational basis state of the qudit. A typical aim of the current method is to calibrate (or optimise) the read-out signal by optimising the one or more read-out optimization parameters.

[0109] A step 606 comprises obtaining a sequence of measurement results by applying a sequence of control signals. The sequence of control signals comprises a repetition of the qudit control signal (as configured in step 602) and the qudit read-out signal (as configured in step 604). Step 606 further comprises measuring the response signals resulting from theapplication of the qudit read-out signals. More in particular, this step may comprise the classical computer sending a sequence of commands to the quantum device which configure the quantum device to apply the qudit control signals and the qudit read-out signals, and to obtain measurements. Thus, in response to sending the control signals to the quantum device, the classical computer may receive a sequence of measurement results from the quantum device.

[0110] The sequence of measurement results can be obtained without initialising the qudit between measuring the response signal and applying the qudit control signal; thus, the ‘output’ state of the qudit after the measurement is (or at least is assumed to be) the ‘input’ state of the next qudit control operation.

[0111] The sequence of measurement results comprises at least k measurements, but more typically comprises at least 20, at least 50, at least 100, at least 200, or at least 500 or more measurements (without intermediate initialisation). The number of measurement results that is used may depend on various criteria, such as an accuracy of the control operations, a spread in the measurement results, an expected occurrence rate of (different types of) errors, time constraints, and / or other criteria.

[0112] In some embodiments, obtaining measurement results comprises alternating a single qudit control signal and a single qudit read-out signal. Other embodiments may comprise additional operations, e.g., additional measurement operations.

[0113] An expected sequence of labels may be determined based on the sequence of control signals, wherein each label corresponds to a computational basis state. The expected sequence of labels is generally symmetric with respect to permutation of the labels.

[0114] A step 608 comprises assigning the measurement results to k measurement bins based on the respective values of the measurement results. The measurement results are typically assigned such that each of the k measurement bins comprises a substantially equal number of measurement results; e.g., each measurement bin may comprise roughly N / k measurement results, wherein N is the total number of measurement results.

[0115] Assigning the measurement results to the k measurement bins may comprise performing a principal component analysis, PCA, on the measurement results and dividing the measurement results into bins along a direction of a principal component. Alternatively, other classification algorithms may be used, e.g., based on clustering. However, principal component analysis is fast, robust, and computationally lightweight.

[0116] A step 610 comprises assigning provisional labels to each of the measurement results in the sequence of measurement results to obtain a sequence of provisional labels. The provisional labels are based on the measurement bin to which a given measurement result has been assigned.A step 612 comprises computing an error based on a deviation between the sequence of provisional labels and an expected sequence of labels. As noted above, the expected sequence of labels is based on the sequence of control signals. In a typical embodiment, a deviation is determined by comparing subsequent provisional labels.

[0117] Typically, computing the error comprises computing a cost function configured to penalize measurement results if the sequence of provisional labels deviates from the expected sequence of labels. For example, such a cost function may be given by

[0118] IW-1

[0119] eR=1 MV — — 71 £ / — I ~mod

[0120]

[0121] n=l

[0122] wherein eRrepresents an error, N represents a number of measurements, mnrepresents a binned measurement result corresponding to measurement number n, and

[0123]

[0124] ..., k - 1} K is a weight function. For example, for k = 3, a typical expected sequence of labels may be 012012012, so that for the expected sequence mn- mn-1= 1 mod k. Thus, the weight function may assign a low or negative weight to input value 1 , and a higher (positive) weight to input values 0 and 2.

[0125] For k = 2, the error function may be given by

[0126] IW-1

[0127] eR=I MV - — 71 £ / — I =mn-l)

[0128]

[0129] n=l

[0130] wherein eRrepresents an error, N represents a number of measurements, mnrepresents a binned measurement result corresponding to measurement number n, and

[0131] f {TRUE, FALSE} -> K is a weight function.

[0132] Other embodiments may use different error functions, e.g., based on a comparison of a provisional label with both the previous and subsequent provisional labels. This may help distinguish between different errors, e.g., errors due to the qudit not being in an expected state, and errors due to assigning an incorrect provisional label. For example, assume that an expected sequence of labels is 0101 and that the corresponding sequence of provisional labels is LHLH. If the wrong provisional label is assigned to the second measurement, the resulting sequence of provisional labels would be LLLH, where the second provisional label is equal to both the previous and subsequent provisional label. However, if the correct provisional label is applied, but the bit-flip fails, the resulting sequence could have been LLHL; and if the measurement induces a bit-flip, the resulting sequence could have been LHHL. In these last two cases, the second provisional label is equal to one neighbour anddifferent from the other neighbour. In practice, these last two cases may be difficult to separate. Some error functions may also take into account statistics based on larger segments or even the entire collection of measurements. For example, due to not all errors being symmetrical, there may be different occurrences of sequences LL than of sequences HH (in an embodiment where alternating sequences are expected).

[0133] A step 614 comprises updating the one or more read-out optimization parameters based on the error.

[0134] A step 616 comprises testing whether a stopping criterion has been met (e.g., a maximum number of iterations, an error smaller than a predetermined threshold, or a decrease in error being smaller than a predetermined threshold). If the stopping criterion has not been met, steps 604-614 may be repeated with the updated one or more read-out optimization parameters, until the stopping criterion has been met, using a known optimisation algorithm.

[0135] In some embodiments, a step 618 comprises, once k clearly separable measurement categories can be distinguished, initialising the qudit in a computational basis state and obtaining additional measurement results. The obtaining additional measurement results can comprise causing the quantum device to read out a state of the qudit, e.g., by applying a read-out pulse and measuring a response signal. These additional measurement results may be used to associate definitive labels with computational basis states, the definitive labels being indicative of the respective computational basis states.

[0136] In some embodiments, the method further comprising updating the one or more qudit-control optimization parameters based on the error, or possibly based on a different error function. In practice, a full calibration may comprise alternately optimising a qudit control signal and a qudit read-out signal, as a more reliable read-out signal allows for finer calibration of the control signal, and vice versa. Similarly, in some cases, different qudit control signals may be applied in different iterations of the described algorithm.

[0137] Fig. 7A depicts a schematic of a hybrid computing system including a classical computer and a quantum device, such as a quantum processing unit (QPU) 714. A quantum algorithm may be defined in terms of quantum circuits which define sequences of quantum gates, which may be regarded as instructions to manipulate the states of the qubits or more general qudits of the quantum device. Quantum gates are typically implemented by applying one or more control signals, e.g., AC signals such as pulses and / or DC signals, to qubits. An operator 702 who wishes to perform a quantum computing experiment, may program an experiment as a sequence of operations 704 (which may be a sequence of quantum gate operations, but also other operations) on a classical computer 706, which may include a CPU, GPU and / or other special purpose processors such as an FPGA. The classical computer may be configured to translate the operations to an instruction set 708 that can beexecuted by a hardware system 710, which may be configured to convert these instructions into analog signals 712 for controlling the qudits which are sent to the QPU. In an embodiment, the hardware system may for example comprise an arbitrary waveform generator (AWG),

[0138] The quantum device may comprise one or more qudits, e.g., qubits, which may be implemented based on any suitable hardware platforms, e.g., superconducting qudits, semiconducting qudits, or qudits based on nitrogen vacancy NV centers in a diamond lattice. After the application of the control signals, the state of the qudits may be measured (usually in response to a read-out signal) resulting in one or more analog measurement signals 716 which are sent back to the classical computer 706. The analog measurement signals may be digitized into a digital measurement signal 720 using a digitizer 718, and processed by the classical computer. Then, the classical computer may either update the hardware instructions based on the processed data 722 (when performing automated experiments), or the operator may interpret the processed data and program a new experiment.

[0139] The shape of the control signals (typically electro-magnetic and / or optical pulses) sent to the qudits of the quantum processor needs to be calibrated by performing a test protocol on the QPU to ensure the control pulses realize the gate operations as defined in the quantum circuit that are executed by the system. The methods described herein allow for fast and accurate calibration of read-out signals.

[0140] Fig. 7B is a block diagram illustrating an exemplary classical data processing system described in this disclosure, for example classical computer 706. Classical data processing system 706 may include at least one processor 752 coupled to memory elements 754 through a system bus 756. As such, the data processing system may store program code within memory elements 754. Further, processor 752 may execute the program code accessed from memory elements 754 via system bus 756. In one aspect, data processing system may be implemented as a computer that is suitable for storing and / or executing program code. It should be appreciated, however, that data processing system 706 may be implemented in the form of any system including a processor and memory that is capable of performing the functions described within this specification.

[0141] Memory elements 754 may include one or more physical memory devices such as, for example, local memory 758 and one or more bulk storage devices 760. Local memory may refer to random access memory or other non-persistent memory device(s) generally used during actual execution of the program code. A bulk storage device may be implemented as a hard drive or other persistent data storage device. The classical data processing system 706 may also include one or more cache memories (not shown) that provide temporary storage of at least some program code in order to reduce the number of times program code must be retrieved from bulk storage device 760 during execution.I nput / output (I / O) devices depicted as key device 762 and output device 764 optionally can be coupled to the data processing system. Examples of key device may include, but are not limited to, for example, a keyboard, a pointing device such as a mouse, or the like.

[0142] Examples of output device may include, but are not limited to, for example, a monitor or display, speakers, or the like. Key device and / or output device may be coupled to data processing system either directly or through intervening I / O controllers. A network adapter 766 may also be coupled to data processing system to enable it to become coupled to other systems, computer systems, remote network devices, and / or remote storage devices through intervening private or public networks. The network adapter may comprise a data receiver for receiving data that is transmitted by said systems, devices and / or networks to said data and a data transmitter for transmitting data to said systems, devices and / or networks. Operation modems, cable operation modems, and Ethernet cards are examples of different types of network adapter that may be used with classical data processing system 706.

[0143] As pictured in FIG. 7B, memory elements 754 may store an application 768. It should be appreciated that classical data processing system 706 may further execute an operating system (not shown) that can facilitate execution of the application. Application, being implemented in the form of executable program code, can be executed by classical data processing system 706, e.g., by processor 752. Responsive to executing application, data processing system may be configured to perform one or more operations to be described herein in further detail.

[0144] In one aspect, for example, classical data processing system 706 may represent a client data processing system. In that case, application 768 may represent a client application that, when executed, configures classical data processing system 706 to perform the various functions described herein with reference to a “client”. Examples of a client can include, but are not limited to, a personal computer, a portable computer, a mobile phone, or the like.

[0145] In another aspect, data processing system may represent a server. For example, data processing system may represent an (HTTP) server in which case application 768, when executed, may configure data processing system to perform (HTTP) server operations. In another aspect, data processing system may represent a module, unit or function as referred to in this specification.

[0146] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” and / or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presenceor addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0147] The corresponding structures, materials, acts, and equivalents of all means or step plus function elements in the claims below are intended to include any structure, material, or act for performing the function in combination with other claimed elements as specifically claimed.

[0148] The description of the present embodiments has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.

Claims

CLAIMS1. A method for calibrating a qudit comprising:configuring (602) a qudit control signal, the qudit control signal being configured to cycle the qudit through k computational basis states associated with the qudit, k being an integer greater than 1 ;configuring (604) a qudit read-out signal based on one or more first optimization parameters, application of the qudit read-out signal to the qudit generating a response signal indicative of a computational basis state of the qudit;obtaining (606) a sequence of measurement results by applying a sequence of control signals, the sequence of control signals comprising a repetition of the qudit control signal and the qudit read-out signal, and measuring the response signals, without initialising the qudit between measuring the response signal and applying the qudit control signal; assigning (608) the measurement results to k measurement bins based on the respective values of the measurement results, each of the k measurement bins comprising a substantially equal number of measurement results;assigning (610) provisional labels to each of the measurement results in the sequence of measurement results to obtain a sequence of provisional labels, the provisional labels being based on the measurement bin to which a given measurement result has been assigned;computing (612) an error based on a deviation between the sequence of provisional labels and an expected sequence of labels, the expected sequence of labels being based on the sequence of control signals; andupdating (614) the one or more first optimization parameters based on the error.

2. The method as claimed in claim 1, wherein the expected sequence of labels is symmetric with respect to permutation of the labels.

3. The method as claimed in claim 1 or 2, wherein assigning the measurement results to the k measurement bins comprises performing a principal component analysis, PCA, on the measurement results and dividing the measurement results into bins along a direction of a principal component.

4. The method as claimed in any one of the preceding claims, wherein obtaining measurement results comprises alternating a single qudit control signal and a single qudit read-out signal.

5. The method as claimed in any one of the preceding claims, wherein computing the error comprises computing a cost function configured to penalize measurement results if the sequence of provisional labels deviates from the expected sequence of labels, preferably the cost function being given byIW-1eR=I Mv - — 71 £ / — I ~mn-l)modn=lwherein eRrepresents an error, N represents a number of measurements, mnrepresents a binned measurement result corresponding to measurement number n, andK is a weight function.

6. The method as claimed in any one of the preceding claims, further comprising repeating the obtaining of the measurement results by repeatedly applying the qudit control signal and the qudit read-out signal and measuring the response signal, the binning of the measurement results, the computing of the error, and the updating of the one or more first optimization parameters until the error meets one or more optimization conditions.

7. The method as claimed in any one of the preceding claims, wherein the qudit control signal is based on one or more second optimization parameters, the method further comprising updating the one or more second optimization parameters based on the error.

8. The method as claimed in any one of the preceding claims, further comprising: initialising the qudit in a known computational basis state;applying the qudit read-out signal, and measuring the response signal; and assigning a definitive label to each measurement bin based on the measured response signal and the known computational basis state.

9. The method as claimed in any one of the preceding claims, wherein the qudit is a qubit, and wherein the qudit control signal is a qubit flip signal configured to flip the qubit from a first computational basis state, e.g. a |0) state, to a second computational basis state different from the first computational basis state, e.g. a |1) state.

10. A hybrid computing system comprising a classical computer (706) and a quantum processor (714), the classical computer (706) comprising at least one processor (752) and a memory (754) storing computer-readable instructions, the memory being communicatively coupled to the at least one processor (752), the classical computer (706) being configured to control the quantum processor (714), the quantum processor (714) comprising a qudit, the atleast one processor (752) being configured to, in response to executing the computer-readable instructions:configure (602) a qudit control signal, the qudit control signal being configured to cycle the qudit through k computational basis states associated with the qudit, k being an integer greater than 1 ;configure (604) a qudit read-out signal based on one or more first optimization parameters, application of the qudit read-out signal to the qudit generating a response signal indicative of a computational basis state of the qudit;obtain (606), using the quantum processor (714), a sequence of measurement results by applying a sequence of control signals, the sequence of control signals comprising a repetition of the qudit control signal and the qudit read-out signal, and measuring the response signals, without initialising the qudit between measuring the response signal and applying the qudit control signal;assign (608) the measurement results to k measurement bins based on the respective values of the measurement results, each of the k measurement bins comprising a substantially equal number of measurement results;assign (610) provisional labels to each of the measurement results in the sequence of measurement results to obtain a sequence of provisional labels, the provisional labels being based on the measurement bin to which a given measurement result has been assigned; compute (612) an error based on a deviation between the sequence of provisional labels and an expected sequence of labels, the expected sequence of labels being based on the sequence of control signals; andupdate (614) the one or more first optimization parameters based on the error.

11. The hybrid computing system as claimed in claim 10, further configured for executing the method of any one of claims 2-9.

12. A computer program product comprising software code portions which, when run by a hybrid computing system comprising a classical computer (706) and a quantum processor (714), configure the hybrid computing system to execute the method steps according to any one of claims 1-9.

13. A non-transitory computer-readable storage medium storing at least one software code portion, the software code portion, when executed or processed by a computer, is configured to perform the method as claimed in any one of claims 1-9.