CN114298317A

CN114298317A - Control parameter determination method and device, computer equipment and storage medium

Info

Publication number: CN114298317A
Application number: CN202111632590.6A
Authority: CN
Inventors: 钱鹏; 夏希德·卡马尔; 肖骁; 刘�东; 顾炎武; 胡孟军; 普亚南
Original assignee: Beijing Institute Of Quantum Information Science; Tsinghua University
Current assignee: Beijing Institute Of Quantum Information Science; Tsinghua University
Priority date: 2021-12-28
Filing date: 2021-12-28
Publication date: 2022-04-08

Abstract

The application relates to a control parameter determination method, a control parameter determination device, a computer device and a storage medium, wherein a standard dependency set is obtained according to basic information of preset quantum bits, a family of proxy models are fitted according to all the dependencies in the standard dependency set, candidate predictive control parameters are determined from multiple groups of predictive control parameters according to the standard dependency set and the family of proxy models, and if the average fidelity calculation value of the candidate predictive control parameters is larger than a preset threshold value, the candidate predictive control parameters are determined as optimal control parameters. The method can accurately and quickly seek the optimal control parameter of the quantum gate, and improves the fidelity of the quantum computer.

Description

Control parameter determination method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of quantum computing, and in particular, to a method and an apparatus for determining a control parameter, a computer device, and a storage medium.

Background

With the development of superconducting quantum chips, quantum computing has entered the era of so-called mesoscale noisy quantum devices.

In general, quantum computers require high fidelity, while high fidelity operation requires calibration of a large number of physical qubits, while calibration of the qubits for tunable coupling requires seeking optimal control parameters for the quantum gates. In the related art, the optimal control parameters are found by a method relying on estimation of a gradient or linear programming.

Then, when seeking the optimal control parameter of the quantum gate in the related art, the accuracy and the speed are not enough, and the fidelity of the quantum computer is reduced.

Disclosure of Invention

In view of the above, it is necessary to provide a control parameter determining method, apparatus, computer device and storage medium for accurately and quickly finding the optimal control parameter of the quantum gate, thereby improving the fidelity of the quantum computer.

In a first aspect, the present application provides a control parameter determining method, including:

acquiring a standard dependency set according to basic information of preset qubits; the standard dependency relationship set comprises a plurality of groups of standard control parameters and the dependency relationship between the corresponding standard quantum gate average fidelity;

fitting a family of proxy models according to all the dependency relationships in the standard dependency relationship set; the family of agent models comprises a plurality of groups of prediction control parameters and the dependency relationship between the corresponding prediction sub-gate average fidelity;

determining candidate predictive control parameters from the plurality of sets of predictive control parameters according to the standard dependency set and the family of proxy models;

and if the average fidelity calculation value of the candidate prediction control parameters is larger than a preset threshold value, determining the candidate prediction control parameters as the optimal control parameters.

In one embodiment, the method further comprises:

and if the average fidelity calculation value of the candidate prediction control parameter is less than or equal to the preset threshold value, adding the dependency relationship between the candidate prediction control parameter and the average fidelity calculation value into the standard dependency relationship set.

In one embodiment, obtaining a standard dependency set according to basic information of a preset qubit includes:

determining multiple groups of standard control parameters according to the basic information of the preset qubits;

determining a pulse waveform corresponding to each standard control parameter according to the basic information of the preset qubit and a plurality of groups of standard control parameters;

and determining the average fidelity of the standard quantum gate corresponding to each standard control parameter according to the pulse waveform corresponding to each standard control parameter to obtain a standard dependency relationship set.

In one embodiment, fitting a family of proxy models to all dependencies in a standard set of dependencies includes:

determining optimization parameters through a preset maximization function according to all the dependency relationships in the standard dependency relationship set; the optimization parameters are corresponding parameters when the function value of the maximization function is maximum;

a family of proxy models is fitted according to the optimization parameters.

In one embodiment, determining candidate predictive control parameters from a plurality of sets of predictive control parameters based on a set of standard dependencies and a family of proxy models comprises:

acquiring a query function between each group of dependency relations in a family of agent models and all dependency relations in a standard dependency relation set;

acquiring a target inquiry function; the target query function represents the query function having the largest function value;

and determining the prediction control parameters in the prediction dependency relationship corresponding to the target inquiry function as candidate prediction control parameters.

In one embodiment, obtaining a query function between each set of dependencies in a family of proxy models and all dependencies in a standard set of dependencies includes:

acquiring statistical parameters between each group of dependency relations in a group of agent models and all dependency relations in a standard dependency relation set, wherein the statistical parameters at least comprise an average value and a covariance;

and acquiring a query function between each group of dependency relations in a family of agent models and all dependency relations in the standard dependency relation set according to the statistical parameters.

In one embodiment, obtaining a target query function comprises:

dividing all the dependency relations in a family of agent models into N relation groups; n is a positive integer;

acquiring the query function with the maximum function value in each relation group to obtain N candidate query functions;

and determining the query function with the maximum function value in the N candidate query functions as the target query function.

In a second aspect, the present application further provides a control parameter determining apparatus, including:

the acquisition module is used for acquiring a standard dependency set according to the basic information of the preset quantum bit; the standard dependency relationship set comprises a plurality of groups of standard control parameters and the dependency relationship between the corresponding standard quantum gate average fidelity;

the fitting module is used for fitting a group of proxy models according to all the dependency relationships in the standard dependency relationship set; the family of agent models comprises a plurality of groups of prediction control parameters and the dependency relationship between the corresponding prediction sub-gate average fidelity;

the first determining module is used for determining candidate prediction control parameters from multiple groups of prediction dependencies according to the standard dependency set and a family of agent models;

and the second determination module is used for determining the candidate prediction control parameters as the optimal control parameters if the average fidelity calculation value of the candidate prediction control parameters is greater than a preset threshold value.

In a third aspect, an embodiment of the present application provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the method provided in any one of the foregoing first aspects when executing the computer program.

In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the method provided in any one of the embodiments in the first aspect.

In a fifth aspect, this application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the computer program implements the steps of the method provided in any one of the embodiments in the first aspect.

According to the control parameter determination method, the control parameter determination device, the computer equipment and the storage medium, a standard dependency relationship set is obtained according to preset basic information of quantum bits, a family of proxy models are fitted according to all dependency relationships in the standard dependency relationship set, candidate prediction control parameters are determined from multiple groups of prediction control parameters according to the standard dependency relationship set and the family of proxy models, and if the average fidelity calculation value of the candidate prediction control parameters is larger than a preset threshold value, the candidate prediction control parameters are determined to be optimal control parameters. In the method, a standard dependence relationship set is obtained through practical experiments, because the standard dependence relationship set comprises the dependence relationship between a plurality of groups of standard control parameters and corresponding standard quantum gate average fidelity, a plurality of groups of prediction control parameters and corresponding prediction quantum gate average fidelity are fitted through the plurality of groups of standard control parameters and the corresponding standard quantum gate average fidelity, candidate prediction control parameters are determined from the plurality of groups of prediction control parameters through a series of calculations, the optimal control parameters are finally determined through comparing the average fidelity calculation values of the candidate prediction control parameters with a preset threshold value, the optimal control parameters can be deduced through comparing a small amount of data information, the efficiency of seeking the optimal control parameters is improved, the optimal values of a group of parameters are not required to be searched independently, the optimal values of the plurality of groups of control parameters are directly sought, the global optimum can be found well, instead of local optimization, the accuracy of finding the optimal control parameter of the quantum gate is improved, so that the fidelity of the quantum computer is improved.

Drawings

FIG. 1 is a diagram of an exemplary implementation of a control parameter determination method;

FIG. 2 is a flow diagram illustrating a method for determining control parameters according to one embodiment;

FIG. 3 is a schematic flow chart of a control parameter determination method according to another embodiment;

FIG. 4 is a schematic flow chart diagram illustrating a control parameter determination method according to another embodiment;

FIG. 5 is a flowchart illustrating a control parameter determination method according to another embodiment;

FIG. 6 is a flowchart illustrating a control parameter determination method according to another embodiment;

FIG. 7 is a flowchart illustrating a control parameter determination method according to another embodiment;

FIG. 8 is a flowchart illustrating a control parameter determination method according to another embodiment;

FIG. 9 is a flowchart illustrating a control parameter determination method according to another embodiment;

FIG. 10 is a block diagram showing the structure of a control parameter determining apparatus according to an embodiment;

FIG. 11 is a diagram illustrating an internal structure of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The control parameter determination method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the quantum device 102 communicates with the server 104 over a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104, or may be located on the cloud or other network server. The quantum device 102 includes a quantum transistor, a quantum memory, a quantum well laser, a quantum interference element sensor, a quantum effect device, a qubit quantum computer, and the like, and the server 104 may be implemented by an independent server or a server cluster composed of a plurality of servers.

The standard dependency set can be obtained from the quantum device 102, the obtained standard dependency set is sent to the server 104, a family of proxy models are fitted in the server 104, and candidate predictive control parameters are searched in the proxy models to determine the optimal control parameters.

In the prior art, when seeking the optimal control parameter of the quantum gate, discrete control parameter points are taken, the gate control is respectively carried out, discrete value points of an objective function related to fidelity are measured, parameter points with better possible effect are deduced by using a linear programming method, or data analysis is carried out to obtain the approximate gradient of the objective function to the control parameter, so that a new control parameter is taken, and then iterative control is carried out to measure the objective function until the objective function, namely the fidelity reaches a relatively optimal value.

However, in the prior art, the parameter search is based on linear programming or gradient-dependent estimation, which is affected by experimental accuracy and errors, and may cause directional deviation of the parameter search and affect final accuracy, and meanwhile, in multi-bit, due to the enlargement of the parameter space, more data needs to be searched to determine the optimized parameters, thereby greatly reducing efficiency.

Based on this, the embodiments of the present application provide a control parameter determination method, apparatus, computer device, and storage medium, which can accurately and quickly seek the optimal control parameter of a quantum gate, and improve the fidelity of a quantum computer.

The following describes in detail the technical solutions of the present application and how the technical solutions of the present application solve the above technical problems by embodiments and with reference to the drawings. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments.

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments.

In one embodiment, as shown in fig. 2, a control parameter determination method is provided, which includes the steps of:

s201, acquiring a standard dependency relationship set according to basic information of preset qubits.

Wherein the standard dependency set comprises dependencies between sets of standard control parameters and corresponding standard quantum gate average fidelity.

A physical quantity is quantized if it has a minimum unit that cannot be continuously divided, and the minimum unit is called quantum, and a bit is the minimum unit representing information, and is information contained in one bit of a binary number or a required amount of information of which 1 is specifically designated from among 2 options.

A bit characterizes the possible states of a signal, and in quantum computing, as a unit of quantum information, is a qubit, which is a quantum system having two independently steerable quantum states (quantumstates), labeled |0 > and |1 >, respectively. Because the quantum bit follows the basic principle of quantum mechanics, the quantum bit can be in any superposition state of the two states according to the state superposition principle of quantum mechanics; the quantum gate is a unitary evolution generated by microwave pulse control, so that the state of the qubit changes.

The basic information of a qubit includes the frequency of the energy range, the bit's dissonance, the height and width of the rabi oscillation, etc.

Alternatively, the qubit may be a superconducting single qubit, which is a qubit implemented by a superconducting quantum circuit, generally consisting of a capacitor and a josephson junction; the determination mode of the basic information of the qubit can be obtained empirically or from a resonant cavity, which is connected to the qubit and is used for reading the information of the qubit.

According to the basic information of the preset qubit, the standard dependency relationship set may be obtained by using a pre-trained neural network model, using the preset qubit as an input of the neural network model, and finally outputting the standard dependency relationship set by training the neural network model.

Optionally, the standard dependency set includes a dependency between a plurality of sets of standard control parameters and corresponding standard quantum gate average fidelity, that is, the standard control parameters and the standard quantum gate average fidelity are in a one-to-one correspondence relationship, and one set of standard control parameters corresponds to one standard quantum gate average fidelity.

S202, fitting a family of proxy models according to all the dependency relationships in the standard dependency relationship set.

The family of proxy models comprises a plurality of groups of prediction control parameters and the dependency relationship between the corresponding prediction quantum gate average fidelity.

The fitting is to know a plurality of data values, and predict corresponding data values according to the known data values; and fitting a group of proxy models according to all the dependency relationships in the standard dependency relationship set by using known multiple groups of standard control parameters and corresponding standard quantum gate average fidelity, and obtaining multiple groups of prediction control parameters and corresponding prediction quantum gate average fidelity through fitting.

According to the method, a group of proxy models are fitted according to all the dependency relationships in the standard dependency relationship set, the fitting mode can be that all the dependency relationships in the standard dependency relationship set are used as input of a preset algorithm through the preset algorithm, multiple groups of prediction control parameters and corresponding prediction sub-gate average fidelity are obtained through the operation of the algorithm, and the dependency relationships between the multiple groups of prediction control parameters and the corresponding prediction sub-gate average fidelity are stored in the proxy models.

And S203, determining candidate predictive control parameters from the multiple groups of predictive control parameters according to the standard dependency relationship set and the family of proxy models.

And determining candidate predictive control parameters from the multiple groups of predictive control parameters according to the multiple groups of standard control parameters and corresponding standard quantum gate average fidelity in the standard dependency relationship set and the multiple groups of predictive control parameters and corresponding predictive quantum gate average fidelity in the family of proxy models.

In one embodiment, the candidate predictive control parameters may be determined from the plurality of sets of predictive dependencies by means of a preset neural network model, using the standard dependency set and a family of proxy models as inputs to the neural network model, and determining the candidate predictive control parameters by training the neural network model, wherein the candidate predictive control parameters are determined from the plurality of sets of predictive control parameters in the proxy model.

In another embodiment, the candidate predictive control parameters may be determined from the plurality of sets of predictive dependencies by using a preset algorithm script, using the standard dependency set and a family of proxy models as the input of the preset algorithm script, and calling the algorithm script to obtain the candidate predictive control parameters.

And S204, if the average fidelity calculation value of the candidate predictive control parameters is larger than a preset threshold value, determining the candidate predictive control parameters as the optimal control parameters.

And determining an average fidelity calculation value corresponding to the candidate predictive control parameter according to the candidate predictive control parameter obtained in the embodiment, and if the average fidelity calculation value corresponding to the candidate predictive control parameter is greater than a preset threshold value, determining the candidate predictive control parameter as an optimal control parameter.

Alternatively, the manner of determining the average fidelity calculation of the candidate predictive control parameter from the candidate predictive control parameter may be through a rabi oscillation experiment.

In one embodiment, the quantum gate takes a single-bit quantum gate as an example, and after obtaining the optimal control parameter, the optimal control parameter can be selected as the control parameter for implementing the single-bit quantum gate, and the calibration process is ended.

In the control parameter determination method, a standard dependency set is obtained according to basic information of preset qubits, a family of proxy models is fitted according to all the dependencies in the standard dependency set, candidate predictive control parameters are determined from multiple groups of predictive control parameters according to the standard dependency set and the family of proxy models, and the candidate predictive control parameters are determined as optimal control parameters if the average fidelity calculation value of the candidate predictive control parameters is greater than a preset threshold value. In the method, a standard dependence relationship set is obtained through practical experiments, because the standard dependence relationship set comprises the dependence relationship between a plurality of groups of standard control parameters and corresponding standard quantum gate average fidelity, a plurality of groups of prediction control parameters and corresponding prediction quantum gate average fidelity are fitted through the plurality of groups of standard control parameters and the corresponding standard quantum gate average fidelity, candidate prediction control parameters are determined from the plurality of groups of prediction control parameters through a series of calculations, the optimal control parameters are finally determined through comparing the average fidelity calculation values of the candidate prediction control parameters with a preset threshold value, the optimal control parameters can be deduced through comparing a small amount of data information, the efficiency of seeking the optimal control parameters is improved, the optimal values of a group of parameters are not required to be searched independently, the optimal values of the plurality of groups of control parameters are directly sought, the global optimum can be found well, instead of local optimization, the accuracy of finding the optimal control parameter of the quantum gate is improved, so that the fidelity of the quantum computer is improved.

The above embodiment describes a case where the average fidelity calculated value of the candidate predictive control parameter is greater than the preset threshold value, and a case where the average fidelity calculated value of the candidate predictive control parameter is less than or equal to the preset threshold value, which is described in detail by an embodiment, which includes: and if the average fidelity calculation value of the candidate prediction control parameter is less than or equal to the preset threshold value, adding the dependency relationship between the candidate prediction control parameter and the average fidelity calculation value into the standard dependency relationship set.

When the average fidelity calculated value of the candidate predictive control parameter is less than or equal to the preset threshold value, and the average fidelity calculated value of the candidate predictive control parameter is the corresponding actually calculated standard quantum gate average fidelity of the candidate predictive control parameter, the dependency between the candidate predictive control parameter and the average fidelity calculated value is added to the standard dependency set.

In one embodiment, as shown in fig. 3, obtaining the standard dependency set according to the basic information of the preset qubits includes the following steps:

s301, determining multiple groups of standard control parameters according to the basic information of the preset qubits.

Optionally, the basic information of the predetermined qubit includes the width of the waveform, the height of the amplitude, B denotes a bottom constant, the duration of the control, the frequency of the microwave, and the dissonance of the qubit. The dissonance of a qubit represents the difference between the 12 and 01 energy levels.

In one embodiment, multiple sets of standard control parameters are determined based on the basic information of the predetermined qubits in such a way that several sets of classical control parameters can be determined as standard control parameters through pre-existing experience.

In another embodiment, the multiple sets of standard control parameters may be determined by randomly selecting values of the multiple sets of control parameters as the standard control parameters.

S302, determining pulse waveforms corresponding to the standard control parameters according to the preset basic information of the qubits and the multiple groups of standard control parameters.

The pulse wave is an electrical signal which is generated suddenly and has a very short discontinuous duration, and the voltage or the current which is generated discontinuously is called pulse voltage or pulse current, and the pulse wave comprises a rectangular wave, a sawtooth wave, a triangular wave, a spike wave, a step wave, a gaussian wave and the like.

In an embodiment, taking an example that a qubit gate is a single-bit qubit gate and a pulse waveform is a gaussian waveform, determining a pulse waveform corresponding to each standard control parameter, where the single-bit qubit gate is controlled by a microwave pulse, and the determining may be performed by generating a time-dependent gaussian waveform pulse by using an arbitrary wave generator according to basic information of qubits preset in the above embodiment and multiple sets of standard control parameters, and the mathematical form of the time-dependent gaussian waveform pulse is represented as:

cos(w_dt)ε_x(t)+sin(w_dt)ε_y(t),t<t_g(1)

wherein t represents time, t_gIndicating the duration of the control, after which the microwave becomes zero, w_dRepresenting the frequency of the microwave,. epsilon_x(t),ε_y(t) is the x and y components of any wave generator, the mathematical form of the x component being:

where σ denotes the width of the waveform, a denotes the amplitude height, and B denotes the bottom constant.

Alternatively, in order to reduce bit leakage caused by the bit being excited to a higher energy level, through analysis of physical theory, the y-component may be set to be a value of a differential adiabatic gate (darg) through analysis of a differential adiabatic gate (a gate)

Wherein the content of the first and second substances,

is epsilon_xThe derivative of (t), δ, is the dissonance of the qubit, i.e., the difference between the 12-level and the 01-level, and this requires previous calibration information, and therefore, in equation (3), the control parameter that needs to be optimized is α. However, setting the y component may result in another phase error.

Therefore, to balance bit leakage and phase errors, a phase factor can be added to the overall waveform to reduce phase errors:

Ω(t)＝ε_x(t)+ε_y(t)→Ω′(t)＝Ω(t)e^i2πγt (4)

where i is an imaginary number, γ is a control parameter, and equation (4) indicates that the microwave driving frequency calibrated before is subjected to fine tuning, and the degree of this fine tuning needs to be optimized. Therefore, the control parameters that need to be optimized are α and γ.

Optionally, the standard control parameters are (α, γ), and the plurality of sets of standard control parameters are a plurality of sets of standard control parameters (α, γ) with different values.

And S303, determining the average fidelity of the standard quantum gate corresponding to each standard control parameter according to the pulse waveform corresponding to each standard control parameter to obtain a standard dependency relationship set.

In one embodiment, the mean fidelity of the standard quantum gate corresponding to each standard control parameter is determined by obtaining a state obtained by the actually controlled gate acting on an eigen state according to a pulse waveform corresponding to the standard control parameter, where the quantum gate is exemplified by a pauli gate, and the mean fidelity of the quantum gate is calculated by the following formula:

wherein, U_ideaRepresenting door operation in theoretical cases, M_pRepresenting a mapping of quantum states in actual conditions, also known as a quantum channel, representing the operation of the gate in non-ideal conditions, p_jA density matrix corresponding to a certain initial state (eigenstate of a Paglix, y and z operator) is represented; the expression (5) means that the difference between the state obtained by the ideal gate acting on the eigenstates of the pauli x, y, z operator and the state obtained by the actually manipulated gate acting on the eigenstates of the pauli x, y, z operator, actually reflects the fidelity of the gate operation.

The pauli x, y, z operators can be represented by three pauli matrices: pauli X matrix, pauli Y matrix, and pauli Z matrix. Wherein the eigenvalues of each pauli matrix are respectively positive and negative 1, the eigenstates are up and down states, and their corresponding eigenvalues are 1 and-1. Wherein the pauli matrices are respectively:

in an embodiment, taking the pauli gate as an example, the states obtained by the action of the actually controlled gate on the eigenstates of the pauli x, y and z operators are determined according to the pulse waveforms corresponding to the standard control parameters, so that the standard quantum gate average fidelity corresponding to the standard control parameters is determined according to the formula (5), and a standard dependency relationship set is obtained, wherein the standard dependency relationship set comprises the standard control parameters and the standard quantum gate average fidelity corresponding to the standard control parameters.

In the standard dependency set of the above embodiments, we obtain the standard quantum gate average fidelity corresponding to each standard control parameter, but in order to reduce the probability of excitation to a higher energy level (leak), there is also an embodiment that obtains a linear combination of the average fidelity and the negative value of leak:

F＝F_g-β*leak (9)

wherein, the beta coefficient and leak are determined by leak experiment, and leak represents the probability of leakage; the standard dependency set now includes the dependencies between each standard control parameter and the corresponding F.

In the control parameter determining method, a plurality of groups of standard control parameters are determined according to the basic information of the preset qubit, the pulse waveform corresponding to each standard control parameter is determined according to the basic information of the preset qubit and the plurality of groups of standard control parameters, and the average fidelity of the standard quantum gate corresponding to each standard control parameter is determined according to the pulse waveform corresponding to each standard control parameter, so that a standard dependency relationship set is obtained. According to the method, the average fidelity of the standard quantum gate corresponding to each standard control parameter is determined according to the pulse waveform corresponding to each standard control parameter, so that a standard dependency relationship set is obtained, a basis is provided for accurately and quickly finding the optimal control parameter of the quantum gate, and the fidelity of a quantum computer is improved.

In one embodiment, as shown in FIG. 4, fitting a family of proxy models from all dependencies in the standard set of dependencies includes:

s401, according to all the dependency relationships in the standard dependency relationship set, determining optimization parameters through a preset maximization function.

The optimization parameters are corresponding parameters when the function value of the maximization function is maximum.

In one embodiment, the optimization parameter is determined by a preset maximization function, where the preset maximization function may be a maximization edge likelihood function, and the maximization edge likelihood function is calculated by:

wherein k is a kernel function of the proxy model, c is a constant value, D is a standard dependency set, and D ═ F_g(α,γ)]And k corresponds to the formula:

k(θ,θ′)＝VR(x)exp(x) (11)

where k (θ, θ ') is called a kernel function of the proxy model (gaussian process), θ and θ' represent the predictive control parameters in two different proxy models, < > represents expectation, V represents a parameter, and r (x) has the following calculation formula:

where x is | θ - θ' |/l.

Therefore, the optimization parameters determined by the maximization function in equation (10) are V and l.

In one embodiment, the method for determining the optimized parameters may be determined by a preset neural network model, and the optimized parameters V and l are finally output by training the neural network model with all the dependencies in the standard dependency set, and the expressions (10) to (12) as inputs of the neural network model.

In another embodiment, the optimization parameter may be determined by replacing equations (11) and (12) in equation (10) according to all the dependencies in the standard dependency set, and calculating the corresponding optimization parameter when the value in equation (10) is the maximum through a preset maximization algorithm.

S402, fitting a family of proxy models according to the optimization parameters.

Optionally, a family of proxy models is fitted according to the obtained optimization parameters, the fitting mode can be obtained in a bayesian mode, and the function value is limited to be between 0 and 1:

f(θ)＝π(g(θ)) (14)

wherein, g (θ) is the proxy model to be obtained, and g (θ) ═ g (θ)₁),...,g(θ_j)]θ is a prediction control parameter (α, γ), and at the same time, this family of proxy models needs to satisfy:

k(θ,θ′)＝<g(θ)g(θ′)>-<g(θ)><g(θ′)> (15)

the predictive control parameters are fitted by determining optimization parameters according to all the dependency relations in the standard dependency set, and the predictive control parameters are mutually consistent with multiple groups of standard control parameters in the standard dependency set.

The proxy model is solved by equations (13) and (14) based on the constraint condition of equation (15).

In one embodiment, a sensitivity value index (Figure of merit, Fom, F for short) is used as a variation curve of the average fidelity of a quantum gate with control parameters to represent the realization capability of the average fidelity of the quantum gate, and the F defines a control scene according to the dependence relationship between the control parameters and the corresponding average fidelity of the quantum gate; according to a Bayesian prediction data mode, a plurality of different control scenes are constructed, a probability is distributed to each control scene, and the probability represents the confidence degree of the scene which is consistent with the reality, namely the average fidelity of a prediction quantum gate corresponding to a prediction control parameter; this provides a complete probability distribution for the control scenario we have preset, which not only allows us to construct a desired control scenario, but also estimates when there is a deviation from the desired behavior.

The agent models represent a guess of the hidden dimensional image models behind the actual data, and the prior probability of the agent models represents the certainty of the guesses, and the certainty changes with the increase of the data.

Predicting the control scene F based on a plurality of groups of standard control parameters and corresponding standard quantum gate average fidelity, finding out the predicted control parameters and the corresponding predicted quantum gate average fidelity, and evaluating the performance of the control scene F through experiments; let F be called a proxy model, representing some feasible estimates and substitutions of our actual scenario.

In the control parameter determining method, all the dependency relationships are concentrated according to the standard dependency relationships, the optimization parameters are determined through the preset maximization function, the optimization parameters are the parameters corresponding to the maximization function when the function value is maximum, and a family of proxy models are fitted according to the optimization parameters. In the method, a group of proxy models are determined by concentrating all the dependency relations through the standard dependency relations, so that the accuracy in searching for the optimal control parameters of the quantum gate is improved.

In one embodiment, as shown in fig. 5, determining candidate predictive control parameters from a plurality of sets of predictive control parameters based on a set of standard dependencies and a family of proxy models includes the steps of:

s501, acquiring a query function between each group of dependency relationships in a family of agent models and all dependency relationships in a standard dependency relationship set.

According to the dependency relationship in a family of agent models and the dependency relationship in a standard dependency relationship set, a query function is determined, the determination mode can be a mode of a neural network model, all the dependency relationships in the agent models and all the dependency relationships in the standard dependency relationship set are used as preset neural network models, and the query function between each group of dependency relationships in the family of agent models and all the dependency relationships in the standard dependency relationship set is obtained by training the neural network.

S502, acquiring a target inquiry function; the target query function indicates the query function having the largest function value.

Optionally, the query function with the largest query function value in the above embodiment is used as the target query function, and the manner of obtaining the target function may be through a maximum value algorithm, where the query function between each group of dependency relationships in the proxy model in the above embodiment and all dependency relationships in the standard dependency relationship set is used as the input of the neural network model, and the query function with the largest function value is determined by calling the maximum value algorithm, and is used as the target query function.

S503, the prediction control parameters in the prediction dependency corresponding to the target inquiry function are determined as candidate prediction control parameters.

Determining a prediction control parameter in a prediction dependency corresponding to a target inquiry function as a candidate prediction control parameter; for example, the prediction control parameter in the prediction dependency corresponding to the target query function is (α)₂,γ₂)，(α₂,γ₂) For the second set of predictive control parameters in the set of predictive control parameters, (. alpha.) is added₂,γ₂) And determining the candidate prediction control parameters.

In the control parameter determination method, a query function between each group of dependency relationships in a group of agent models and all dependency relationships in a standard dependency relationship set is obtained; acquiring a target inquiry function; the target query function represents the query function having the largest function value; and determining the prediction control parameters in the prediction dependency relationship corresponding to the target inquiry function as candidate prediction control parameters. According to the method, the prediction control parameter with the maximum corresponding value of the query function is determined through the query function, the prediction control parameter is determined as the candidate prediction control parameter, the candidate control parameter is determined in the prediction control parameter group, the efficiency of searching for the optimal control parameter is improved, and when the optimal value of the control parameter is searched, the optimal control parameter is directly searched in multiple groups of prediction control parameters, so that the accuracy of searching for the optimal control parameter is improved.

In one embodiment, as shown in fig. 6, obtaining a query function between each set of dependencies in a family of proxy models and all dependencies in a standard set of dependencies includes the following steps:

s601, obtaining statistical parameters between each group of dependency relations in a group of agent models and all dependency relations in a standard dependency relation set.

Wherein the statistical parameters include at least a mean and a covariance.

In one embodiment, the average value between each set of dependencies in a family of proxy models and all dependencies in a standard set of dependencies is:

wherein the content of the first and second substances,

denotes the average value of f (theta, phi) under the prediction control parameters (theta, phi), k is the kernel function in the above embodiment, D is the standard dependency set, theta is the prediction control parameter, theta is_jAnd theta_jThe parameters are controlled for different criteria in different criteria dependencies.

The covariance between each set of dependencies in a family of proxy models and all dependencies in the standard dependency set is:

cov(f)＝k(θ,θ)-k(θ,θ_i)k^-1(θ_i,θ_j)k((θ_i,θ) (17)

where k is the kernel function in the above embodiment.

In one embodiment, after determining the proxy model, a probability of generation p (f) may be determined, which is:

where N (a, b) represents the probability density of a Gaussian distribution with mean a and covariance b.

S602, according to the statistical parameters, acquiring a query function between each group of dependency relationships in a family of agent models and all dependency relationships in a standard dependency relationship set.

In one embodiment, the mean and covariance in the above embodiment are used to design the query function as:

wherein mu is a parameter factor and can be determined through historical experience, and the size of mu determines the convergence rate and the local size. The larger mu is, the less likely to fall into a local extremum, and the smaller mu is, the faster the convergence is. We can dynamically adjust μ to take a relatively large value during the initial search and start to slowly decay to a small value later.

In the control parameter determination method, statistical parameters between each group of dependency relationships in a group of proxy models and all dependency relationships in a standard dependency relationship set are obtained, wherein the statistical parameters at least comprise an average value and a covariance; and acquiring a query function between each group of dependency relations in a family of agent models and all dependency relations in the standard dependency relation set according to the statistical parameters. The method can accurately and quickly seek the optimal control parameter of the quantum gate, and improves the fidelity of the quantum computer.

In one embodiment, obtaining a target query function comprises the steps of:

s701, dividing all the dependency relations in a family of agent models into N relation groups.

Wherein N is a positive integer.

For example, if a family of agent models includes 12 dependencies, the 12 dependencies are divided into N relationship groups, where N may be 3, then the 12 dependencies are divided into 3 relationship groups.

In an embodiment, the relationship groups may be divided by a preset division algorithm, and all the dependencies and N in a group of proxy models are used as inputs of the preset division algorithm, and the algorithm is run to obtain N divided relationship groups.

Optionally, in practical applications, the value of N is not limited.

S702, the query function with the maximum function value in each relation group is obtained, and N candidate query functions are obtained.

In one embodiment, the query function with the largest value is found in each relationship group according to the maximum value method, and the query function with the largest value corresponding to each relationship group is used as the candidate query function, so that N relationship groups obtain N candidate query functions.

S703, determine the query function with the largest function value among the N candidate query functions as the target query function.

In one embodiment, the query function with the largest value among the N candidate query functions is used as the target query function, for example, there are 4 candidate query functions including candidate query functions a, b, c, and d, and if the value corresponding to the query function a is the largest among the 4 query functions, the query function a is used as the target query function.

In the control parameter determining method, all the dependency relationships in a family of proxy models are divided into N relationship groups, where N is a positive integer, the query function with the largest function value in each relationship group is obtained, N candidate query functions are obtained, and the query function with the largest function value in the N candidate query functions is determined as the target query function. In the method, the agent model data is learned according to a Bayesian mode, an optimal target query function is searched, the optimal control parameters of the quantum gate can be accurately and quickly searched, and the fidelity of the quantum computer is improved.

In one embodiment, as shown in FIG. 8, FIG. 8 is a flow chart of a control parameter determination method.

In one embodiment, the control parameters are optimized, and in an experiment, the number of searches is generally 10 orders of magnitude for the accuracy of the control parameters of the experiment. Typically two control parameters are experimentally 0.01 and Mhz accuracy, while the search domains are roughly [ -1,0] and [ -100,0] Mhz, the total parameter search point is of the order of 1e5, since the strategy of differential adiabatic gates is a split search, roughly requiring the square of the number of single parameter searches. In the embodiment, the optimal value is directly found according to the two-dimensional parameter set, only a few data points are needed, and the efficiency is improved in experimental resources and time.

The embodiment of learning the optimal control parameter by statistics on some data by using the bayesian method has the advantages that: the method is insensitive to errors, disturbance and noise based on a statistical method; by comparing less information of a part of data, the optimal control parameter can be deduced, and the efficiency is improved; global optimization can be well found, but not local optimization; the optimization of the high-dimensional parameter group is directly found without independently searching the optimized value of one parameter, so that the searching efficiency is improved.

In one embodiment, a proxy model is fitted through actual control parameters and corresponding fidelity, the proxy model comprises predicted control parameters matched with the actual control parameters and corresponding predicted average fidelity, corresponding candidate control parameters in the proxy model are obtained according to numerical calculation, and real experiments are carried out on the candidate control parameters to obtain the real average fidelity corresponding to the candidate control parameters.

Then, comparing the real average fidelity corresponding to the candidate control parameter with a preset fidelity threshold, and if the real average fidelity is greater than or equal to the preset fidelity threshold, determining the candidate control parameter as the optimal control parameter; and if the fidelity is smaller than the preset fidelity threshold, adding the candidate control parameters and the corresponding real average fidelity into the data sets of the actual control parameters and the corresponding fidelity, re-fitting the proxy parameters, and continuously executing the steps.

In one embodiment, as shown in fig. 9, this embodiment includes the steps of:

and S901, determining basic information of the qubits, including frequency, height, control duration, bit dissonance and the like of the energy range.

And S902, setting a plurality of groups of standard control parameters and corresponding pulse waveforms according to the basic information of the qubits.

And S903, obtaining the average fidelity of the quantum gates corresponding to the multiple groups of standard control parameters according to the multiple groups of pulse waveforms.

And S904, fitting a group of proxy models according to the multiple groups of standard control parameters and the corresponding average fidelity, wherein the proxy models comprise prediction control parameters and prediction fidelity.

S905, determining the average value and the covariance under the standard control parameters and the prediction control parameters in the proxy model.

S906, determining a query function according to the average value and the covariance, calculating the maximum value in the query function, and determining the predictive control parameter corresponding to the maximum value.

And S907, determining the average fidelity of the quantum gate corresponding to the predictive control parameter in the S906.

S908, judging whether the average fidelity in S907 is larger than a fidelity threshold, if so, outputting corresponding control parameters, namely optimal control parameters; otherwise, adding the predicted control parameter and the average fidelity of the corresponding quantum gate into the standard control parameter and the average fidelity of the corresponding quantum gate, and continuing to execute S905-S907.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the present application further provides a control parameter determining apparatus for implementing the above-mentioned control parameter determining method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so specific limitations in one or more embodiments of the control parameter determining device provided below can be referred to the limitations of the control parameter determining method in the foregoing, and details are not described herein again.

In one embodiment, as shown in fig. 10, there is provided a control parameter determination apparatus 1000, including an obtaining module 1001, a fitting module 1002, a first determining module 1003, and a second determining module 1004, wherein:

an obtaining module 1001, configured to obtain a standard dependency set according to basic information of a preset qubit; the standard dependency relationship set comprises a plurality of groups of standard control parameters and the dependency relationship between the corresponding standard quantum gate average fidelity;

a fitting module 1002, configured to fit a family of proxy models according to all dependency relationships in the standard dependency relationship set; the family of agent models comprises a plurality of groups of prediction control parameters and the dependency relationship between the corresponding prediction sub-gate average fidelity;

a first determining module 1003, configured to determine candidate prediction control parameters from multiple sets of prediction dependencies according to the standard dependency set and the family of proxy models;

a second determining module 1004, configured to determine the candidate predictive control parameter as the optimal control parameter if the average fidelity calculation value of the candidate predictive control parameter is greater than the preset threshold.

In one embodiment, the apparatus 1000 further comprises:

and the adding module is used for adding the dependency relationship between the candidate prediction control parameters and the average fidelity calculated value into the standard dependency relationship set if the average fidelity calculated value of the candidate prediction control parameters is less than or equal to a preset threshold value.

In one embodiment, the obtaining module 1001 includes:

the first determining unit is used for determining a plurality of groups of standard control parameters according to the basic information of the preset quantum bit;

the second determining unit is used for determining the pulse waveform corresponding to each standard control parameter according to the preset basic information of the qubits and the multiple groups of standard control parameters;

and the third determining unit is used for determining the average fidelity of the standard quantum gate corresponding to each standard control parameter according to the pulse waveform corresponding to each standard control parameter to obtain a standard dependency relationship set.

In one embodiment, fitting module 1002 includes:

the fourth determining unit is used for determining the optimization parameters through a preset maximization function according to all the dependency relationships in the standard dependency relationship set; the optimization parameters are corresponding parameters when the function value of the maximization function is maximum;

and the fitting unit is used for fitting a family of proxy models according to the optimization parameters.

In one embodiment, the first determining module 1003 includes:

the first acquisition unit is used for acquiring a query function between each group of dependency relationships in a family of agent models and all dependency relationships in a standard dependency relationship set;

a second acquisition unit configured to acquire a target query function; the target query function represents the query function having the largest function value;

and the fifth determining unit is used for determining the prediction control parameters in the prediction dependency relationship corresponding to the target inquiry function as candidate prediction control parameters.

In one embodiment, the first obtaining unit includes:

the first obtaining subunit is used for obtaining statistical parameters between each group of dependency relationships in a group of agent models and all dependency relationships in a standard dependency relationship set, wherein the statistical parameters at least comprise an average value and a covariance;

and the second acquiring subunit is used for acquiring a query function between each group of dependency relationships in the group of proxy models and all dependency relationships in the standard dependency relationship set according to the statistical parameters.

In one embodiment, the second obtaining unit includes:

the dividing subunit is used for dividing all the dependency relationships in a family of agent models into N relationship groups; n is a positive integer;

the obtaining subunit is used for obtaining the query function with the maximum function value in each relationship group to obtain N candidate query functions;

and the determining subunit is used for determining the query function with the largest function value in the N candidate query functions as the target query function.

The various modules in the control parameter determination device described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing control parameter determination data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a control parameter determination method.

Those skilled in the art will appreciate that the architecture shown in fig. 11 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the following steps when executing the computer program:

In one embodiment, the processor, when executing the computer program, performs the steps of:

a family of proxy models is fitted according to the optimization parameters.

In the steps implemented by the processor in this embodiment, the implementation principle and technical effect are similar to those of the control parameter determination method described above, and are not described herein again.

In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

In the present embodiment, the implementation principle and technical effect of each step implemented when the computer program is executed by the processor are similar to the principle of the control parameter determination method described above, and are not described herein again.

In an embodiment, a computer program product is provided, comprising a computer program which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), Magnetic Random Access Memory (MRAM), Ferroelectric Random Access Memory (FRAM), Phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A control parameter determination method, the method comprising:

acquiring a standard dependency set according to basic information of preset qubits; the standard dependency relationship set comprises dependency relationships between a plurality of groups of standard control parameters and corresponding standard quantum gate average fidelity;

fitting a family of proxy models according to all the dependency relationships in the standard dependency relationship set; the family of agent models comprises a dependency relationship between a plurality of groups of predictive control parameters and corresponding predictive quantum gate average fidelity;

determining candidate predictive control parameters from the multiple sets of predictive control parameters according to the standard dependency set and the family of proxy models;

and if the average fidelity calculation value of the candidate prediction control parameters is larger than a preset threshold value, determining the candidate prediction control parameters as optimal control parameters.

2. The method of claim 1, further comprising:

3. The method according to claim 1 or 2, wherein the obtaining a standard dependency set according to the basic information of the predetermined qubits comprises:

determining a pulse waveform corresponding to each standard control parameter according to the basic information of the preset qubit and the multiple groups of standard control parameters;

and determining the average fidelity of the standard quantum gate corresponding to each standard control parameter according to the pulse waveform corresponding to each standard control parameter to obtain the standard dependency relationship set.

4. The method of claim or the method, wherein fitting a family of proxy models to all dependencies in the standard set of dependencies comprises:

determining an optimization parameter through a preset maximization function according to all the dependency relations in the standard dependency relation set; the optimization parameter is a parameter corresponding to the maximum function value of the maximization function;

and fitting the family of proxy models according to the optimization parameters.

5. The method of claim 1 or 2, wherein determining candidate predictive control parameters from the plurality of sets of predictive control parameters according to the set of standard dependencies and the family of surrogate models comprises:

acquiring a query function between each group of dependency relations in the group of agent models and all dependency relations in the standard dependency relation set;

acquiring a target inquiry function; the target inquiry function represents an inquiry function with the maximum function value;

and determining the prediction control parameters in the prediction dependency relationship corresponding to the target inquiry function as the candidate prediction control parameters.

6. The method of claim 5, wherein obtaining a query function between each set of dependencies in the family of agent models and all dependencies in the set of standard dependencies comprises:

obtaining statistical parameters between each group of dependency relations in the group of agent models and all dependency relations in the standard dependency relation set, wherein the statistical parameters at least comprise an average value and a covariance;

and acquiring a query function between each group of dependency relations in the group of agent models and all dependency relations in the standard dependency relation set according to the statistical parameters.

7. The method of claim 5, wherein obtaining the target query function comprises:

dividing all the dependency relations in the family of agent models into N relation groups; n is a positive integer;

8. A control parameter determination apparatus, characterized in that the apparatus comprises:

the acquisition module is used for acquiring a standard dependency set according to the basic information of the preset quantum bit; the standard dependency relationship set comprises dependency relationships between a plurality of groups of standard control parameters and corresponding standard quantum gate average fidelity;

the fitting module is used for fitting a group of proxy models according to all the dependency relationships in the standard dependency relationship set; the family of agent models comprises a dependency relationship between a plurality of groups of predictive control parameters and corresponding predictive quantum gate average fidelity;

a first determining module, configured to determine candidate predictive control parameters from the multiple sets of predictive dependencies according to the standard dependency set and the family of proxy models;

and the second determination module is used for determining the candidate prediction control parameter as the optimal control parameter if the average fidelity calculation value of the candidate prediction control parameter is greater than a preset threshold value.

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.