KR20240018909A - Apparatus and method for classifying quantum states in quantum information integrated processing system - Google Patents

Abstract

The present invention relates to a quantum state classification device and method in a quantum information integrated processing system, and to a technique for classifying qubit states at the pulse level by training a machine learning-based classifier model and testing the trained classifier model, A quantum state classification device according to an embodiment of the present invention includes a data set generator that generates a training data set and an evaluation data set related to the qubit state, and trains a plurality of classifiers using the generated training data set. a classifier training unit, which calculates the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the generated evaluation data set, and calculates the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the calculated classification accuracy; A classifier evaluation unit that evaluates the classifier, and classifies the qubit state into an entangled state and a separable state using any one of the plurality of classifiers based on the calculated classification accuracy. It may include a quantum state classification unit.

Description

Quantum state classification device and method in quantum information integrated processing system {APPARATUS AND METHOD FOR CLASSIFYING QUANTUM STATES IN QUANTUM INFORMATION INTEGRATED PROCESSING SYSTEM}

The present invention relates to a quantum state classification device and method in a quantum information integrated processing system, and to a technique for classifying qubit states at the pulse level by training a machine learning-based classifier model and testing the trained classifier model. The quantum state classification device and method can be applied to the quantum state measurement processing unit of the quantum observation and analysis platform in the quantum information technology integrated system.

Quantum computing is a rapidly emerging technology that leverages unprecedented resources to solve problems too complex for conventional computers.

Quantum computers use properties of quantum physics to store data and perform calculations.

Using quantum superposition and entanglement phenomena, quantum computers can solve a large number of complex problems.

Different possible input combinations can be used simultaneously and this is where it has an advantage over classical computers.

Quantum computers are made up of qubits, and the most popular qubit approach involves superconducting qubits, which are being developed by IBM, Google, Regetti, and Intel.

IBM is considered one of the leaders in the quantum field, and its quantum computers provide researchers with cloud access.

Recently, IBM gave a small number of quantum computers access to the highest level of control with the help of Qiskit Pulse, a way to interact with quantum computers with pulses.

Any quantum process, such as running a quantum circuit or algorithm, involves three steps: (i) state preparation and initialization, (ii) state evolution, and (iii) state estimation and measurement.

The output of all quantum circuits is probabilistic, and the process of determining this is called state estimation and measurement.

For IBM at the hardware level, a machine learning-based classifier determines the state of a qubit.

A qubit is a bit that can be either 0 or 1 in a classical computer, while a quantum computer has a qubit that is represented by a |ψ> state, and the interesting thing is that it can be in a superposition of 0 and 1, and a qubit is It is written as |ψ> = α |0> + β|1>.

where α and β are probability amplitudes and the sum of squares is 1, when reading a qubit the quantum state decays to a ground state |0> with probability ||α||2 and an excited state |1> with probability ||β||2 do.

After all quantum calculations, these probability amplitudes are determined, and this process is called qubit state estimation.

IBM's superconducting quantum computer uses a specific type of superconducting qubit, and in the energy spectrum, the ground energy level is denoted as |0> and the excited energy level is denoted as |1>.

Qiskit Pulse Like classical programs, quantum programs go through a compilation process from high-level programming to assembly language.

However, unlike classical, quantum hardware is controlled by analog pulses, and the four steps of a quantum compiler can be expressed as (i) programming language, (ii) assembly language, (iii) gate, and (iv) pulse schedule.

Extracting the best performance from a quantum computer requires pulse-level control that cannot be achieved with standard circuit models, and Qiskit Pulse, a Python kit, provides a precise and was developed well.

Pulse coding can help explore different ways to reduce noise, apply dynamic decoupling, and achieve optimal control in quantum computers.

For example, a Gaussian pulse of a certain amplitude and duration can be sent to move a qubit from the ground state |0> to the excited state |1> or vice versa.

As entanglement is an essential resource that exhibits quantum advantages in quantum information processing tasks, it is important to reliably detect entanglement in quantum states.

A quantum information integration system is needed that can support not only quantum state measurement processing, but also various quantum information processing to support the implementation and research of related technologies.

Korean Patent Publication No. 10-2022-0047128, “Quantum circuit error correction method and device” Korean Patent Publication No. 10-2022-0040460, “Pipelined hardware decoder for quantum computing devices” U.S. Patent Publication No. 2021/0398007, “QUANTUM PROCESSING SYSTEM” Korean Patent Publication No. 10-2021-0084222, “Ray classification device and method based on quantized convergence direction”

The purpose of the present invention is to provide a quantum state classification device and method for classifying qubit states at the pulse level by training a machine learning-based classifier model and testing the trained classifier model.

The purpose of the present invention is to improve the performance of quantum state classification by evaluating the performance of various machine learning-based classifiers in hardware for qubit state estimation and selecting classifiers with excellent performance.

The present invention aims to provide a Python-based platform design for calculating, plotting and comparing various machine learning basic classifiers for qubit reading demonstrated on IBM quantum computers.

The present invention is an artificial neural network (ANN)-based classifier capable of separating entanglements for arbitrary bipartite states based on traditional entanglement detection methods such as Bell or Clauser-Horne-Shimony-Holt (CHSH) inequality. The purpose is to provide.

A quantum state classification device according to an embodiment of the present invention includes a data set generator that generates a training data set and an evaluation data set related to the qubit state, and trains a plurality of classifiers using the generated training data set. a classifier training unit, which calculates the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the generated evaluation data set, and calculates the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the calculated classification accuracy; A classifier evaluation unit that evaluates the classifier, and classifies the qubit state into an entangled state and a separable state using any one of the plurality of classifiers based on the calculated classification accuracy. It may include a quantum state classification unit.

The data set generator estimates a ground state, an excited state, and a mixed state for the qubit state based on pulse level control, and the estimated ground state The training data set and the evaluation data set can be generated in a one-to-one ratio by randomly shuffling the ground state, excited state, and mixed state.

The plurality of classifiers include a linear discriminant analysis (LDA)-based classifier, a support vector machine (SVM)-based classifier, a k-nearest neighbor (KNN)-based classifier, a neural network (NNet)-based classifier, and an artificial neural network (ANN)-based classifier. may include.

The classifier training unit can train the ANN (artificial neural network)-based classifier using Bell or CHSH (Clauser-Horne-Shimony-Holt) inequality values as feature vectors of the generated training data set. there is.

The feature vector is partial information of the qubit and is calculated as a weight by an artificial neural network, and the range of entangled state detection of the Bell operator or CHSH operator can be generalized to any state.

The artificial neural network is constructed using either linear optimization or non-linear optimization, and when constructed using non-linear optimization, classification accuracy may increase as the number of neurons in the hidden layer increases.

The classifier evaluation unit determines the performance of the classifier by calculating the median and the interquartile range in relation to the classification accuracy for classifying qubit read data at the pulse level for each of the plurality of classifiers, and determines the performance of the determined classifier. The plurality of classifiers can be evaluated by comparing them with each other.

The quantum state classification method according to an embodiment of the present invention includes the steps of generating a training data set and an evaluation data set related to the qubit state in a data set generating unit, and generating the generated training data set in a classifier training unit. A step of training a plurality of classifiers using a classifier evaluation unit, calculating the median and interquartile range of classification accuracy for the trained plurality of classifiers based on the generated evaluation data set, and , evaluating the plurality of classifiers based on the calculated classification accuracy, and in a quantum state classification unit, converting the qubit state into an entangled state using any one of the plurality of classifiers based on the calculated classification accuracy. It may include the step of classifying into an entangled state and a separable state.

The step of generating a training data set and an evaluation data set related to the Qubit state includes a ground state for the Qubit state based on pulse level control, here Estimating an excited state and a mixed state, randomly shuffling the estimated ground state, excited state, and mixed state in a one-to-one ratio to form the training data set and the evaluation. It may include creating a data set.

The plurality of classifiers include a linear discriminant analysis (LDA)-based classifier, a support vector machine (SVM)-based classifier, a k-nearest neighbor (KNN)-based classifier, a neural network (NNet)-based classifier, and an artificial neural network (ANN)-based classifier. may include.

The step of training a plurality of classifiers using the generated training data set uses Bell or CHSH (Clauser-Horne-Shimony-Holt) inequality values as feature vectors of the generated training data set. This may include training the ANN (artificial neural network)-based classifier.

Calculate the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the generated evaluation data set, and evaluate the plurality of classifiers based on the calculated classification accuracy. The step of determining the performance of the classifier by calculating the median and the interquartile range in relation to the classification accuracy for classifying qubit read data at the pulse level for each of the plurality of classifiers, and determining the performance of the determined classifier. It may include evaluating the plurality of classifiers by comparing them with each other.

The present invention can provide a quantum state classification device and method for classifying qubit states at the pulse level by training a machine learning-based classifier model and testing the trained classifier model.

The present invention can improve the performance of quantum state classification by evaluating the performance of various machine learning-based classifiers in hardware for qubit state estimation and selecting a classifier with excellent performance.

The present invention can provide a Python-based platform design for calculating, plotting and comparing various machine learning base classifiers for qubit reading demonstrated on IBM quantum computers.

The present invention is an artificial neural network (ANN)-based classifier capable of separating entanglements for arbitrary bipartite states based on traditional entanglement detection methods such as Bell or Clauser-Horne-Shimony-Holt (CHSH) inequality. It can be provided as .

1 is a diagram illustrating a quantum state classification device according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating an IQ plot for states for generating a data set in a quantum state classification device according to an embodiment of the present invention.
Figure 3 is a diagram illustrating the accuracy evaluation results by the classifier evaluation unit in the quantum state classification device according to an embodiment of the present invention.
FIGS. 4A and 4B are diagrams illustrating an artificial neural network (ANN) model trained by a classifier training unit in a quantum state classification device according to an embodiment of the present invention.
Figure 4c is a diagram explaining the accuracy evaluation results for the ANN model according to an embodiment of the present invention.
Figure 5 is a diagram explaining a quantum state classification method according to an embodiment of the present invention.
Figure 6 is a diagram for explaining a counterfactual entanglement distribution model according to an embodiment of the present invention.
Figure 7 is a diagram showing the concept of controlled quantum teleportation according to an embodiment of the present invention.
Figure 8 is a flowchart showing the operation of a controlled counterfactual quantum teleportation support device according to an embodiment of the present invention.
Figure 9 is a diagram explaining a mechanism for performing QST in qubit transmission.
Figure 10 shows the Log infidelity of the reconstructed state and the actual state for pure states (a) and mixed states (b).
Figure 11 is a block diagram of a tomographic imaging device according to an embodiment of the present invention.
Figure 12 is a diagram showing a quantum information technology integrated system according to an embodiment of the present invention.

Hereinafter, various embodiments of this document are described with reference to the attached drawings.

The embodiments and terms used herein are not intended to limit the technology described in this document to a specific embodiment, and should be understood to include various changes, equivalents, and/or substitutes for the embodiments.

In the following description of various embodiments, if a detailed description of a related known function or configuration is judged to unnecessarily obscure the gist of the invention, the detailed description will be omitted.

The terms described below are terms defined in consideration of functions in various embodiments, and may vary depending on the intention or custom of the user or operator. Therefore, the definition should be made based on the contents throughout this specification.

In connection with the description of the drawings, similar reference numbers may be used for similar components.

Singular expressions may include plural expressions, unless the context clearly indicates otherwise.

In this document, expressions such as “A or B” or “at least one of A and/or B” may include all possible combinations of the items listed together.

Expressions such as “first,” “second,” “first,” or “second,” can modify the corresponding components regardless of order or importance and are used to distinguish one component from another. It is only used and does not limit the corresponding components.

When a component (e.g. a first) component is said to be "connected (functionally or communicatively)" or "connected" to another (e.g. a second) component, it means that the component is connected to the other component. It may be connected directly to a component or may be connected through another component (e.g., a third component).

In this specification, “configured to” means “suitable for,” “having the ability to,” or “changed to,” depending on the situation, for example, in terms of hardware or software. ," can be used interchangeably with "made to," "capable of," or "designed to."

In some contexts, the expression “a device configured to” may mean that the device is “capable of” working with other devices or components.

For example, the phrase "processor configured (or set) to perform A, B, and C" refers to a processor dedicated to performing the operations (e.g., an embedded processor), or by executing one or more software programs stored on a memory device. , may refer to a general-purpose processor (e.g., CPU or application processor) capable of performing the corresponding operations.

Additionally, the term 'or' means 'inclusive or' rather than 'exclusive or'.

That is, unless otherwise stated or clear from the context, the expression 'x uses a or b' means any of the natural inclusive permutations.

Terms such as '..unit' and '..unit' used hereinafter refer to a unit that processes at least one function or operation, and may be implemented as hardware, software, or a combination of hardware and software.

1 is a diagram illustrating a quantum state classification device according to an embodiment of the present invention.

1 illustrates the components of a quantum state classification device according to an embodiment of the present invention.

Referring to FIG. 1, the quantum state classification device 100 according to an embodiment of the present invention includes a data set generation unit 110, a classifier training unit 120, a classifier evaluation unit 130, and a quantum state classification unit 140. Includes.

For example, the data set generator 110 generates a training data set and an evaluation data set related to the qubit state.

According to one embodiment of the present invention, the data set generator 110 measures the raw in-phase and quadrature output values of the qubit in the ground state and excited state in relation to the labeled data set to set a training data set and an evaluation data set. can be created.

For example, the data set generator 110 estimates the ground state, excited state, and mixed state of the qubit state based on pulse level control.

In addition, the data set generator 110 randomly shuffles the estimated ground state, excited state, and mixed state to generate a training data set and an evaluation data set in a one-to-one ratio. You can.

According to one embodiment of the present invention, the data set generator 110 prepares the qubit in the ground state and uses the pulse level (level 1 measurement) using Qiskit Pulse without performing any other work. Measure the qubit.

Next, the data set generator 110 secures in-phase and quadrature components of the output wave that serves as a shape data set based on the measured qubits.

Afterwards, the data set generator 110 puts the qubit into an excited state and generates in-phase and quadrature of the output wave that serves as a shape data set based on the measured qubit. Secure the components.

Accordingly, the data set generator 110 secures 1024 shots for each of the ground state and the excited state, securing a total of 2048 shots, and after random shuffling, 50% is used as a training data set and 50% is used as test data. Created as a set.

According to one embodiment of the present invention, the classifier training unit 120 trains a plurality of classifiers using the training data set generated by the data set generating unit 110.

For example, multiple classifiers include a linear discriminant analysis (LDA)-based classifier, a support vector machine (SVM)-based classifier, a k-nearest neighbor (KNN)-based classifier, a neural network (NNet)-based classifier, and an artificial neural network (ANN). May include a base classifier.

For example, the classifier training unit 120 may train an ANN-based classifier using Bell or Clauser-Horne-Shimony-Holt (CHSH) inequality values as feature vectors of the training data set.

For example, the feature vector is partial information of the qubit and is calculated as a weight by an artificial neural network, and the range of entangled state detection of the Bell operator or CHSH operator can be generalized to any state. there is.

For example, an artificial neural network is constructed using either linear optimization or non-linear optimization, and when constructed using non-linear optimization, classification accuracy may increase as the number of neurons in the hidden layer increases.

According to one embodiment of the present invention, the classifier evaluation unit 130 calculates the median and interquartile range of classification accuracy for a plurality of classifiers trained based on the data set, and calculates the calculated classification accuracy. Based on this, multiple classifiers can be evaluated.

For example, once the classifier is trained, the classifier evaluation unit 130 evaluates the performance of the machine learning method using accuracy as a measurement standard.

The classifier evaluation unit 130 calculates the median and interquartile accuracy range for each of the plurality of classifiers to obtain an approximation of the ability to accurately classify qubit read data at the pulse level.

For example, the classifier evaluation unit 130 calculates the highest accuracy score for the classifier with the highest median value and the smallest interquartile range.

The median value represents average performance, and the interquartile range is related to the range of highest and lowest accuracy, so the combination of median and interquartile range can be used to judge the performance of the classifier.

According to one embodiment of the present invention, the classifier evaluation unit 130 determines the performance of the classifiers by calculating the median and interquartile range in relation to the classification accuracy of classifying qubit read data at the pulse level for each of the plurality of classifiers. , multiple classifiers can be evaluated by comparing the performances of the determined classifiers.

According to one embodiment of the present invention, the quantum state classification unit 140 uses one of a plurality of classifiers based on the classification accuracy calculated by the classifier evaluation unit 130 to classify the qubit state into an entangled state. ) and separable states.

That is, the quantum state classification unit 140 evaluates among a plurality of machine learning model-based classifiers, selects a classifier with relatively excellent performance based on the evaluation results, and classifies the quantum state based on the selected classifier. In doing so, it improves the inaccuracies associated with the classification of existing quantum states and the associated shortcomings.

Therefore, the present invention can provide a quantum state classification device and method for classifying qubit states at the pulse level by training a machine learning-based classifier model and testing the trained classifier model.

FIG. 2 is a diagram illustrating an IQ plot for states for generating a data set in a quantum state classification device according to an embodiment of the present invention.

Referring to FIG. 2, graph 200 shows Q on the horizontal axis and I on the vertical axis, which is 1024 measuring the ground state |0> and the excited state |1> using measurement level 1. It could be an IQ plot of a dog's trail.

This is done on the IBM quantum computer using Qiskit Pulse, the qubits are first prepared in the default state |0>, which is also the default state of the IBM quantum computer, and without doing anything else, use the measurement level 1 setup to change the state. Measure.

I get complex values from Qiskit Pulse. This provides functionality for each qubit state.

This is done over 1024 trails to generate a ground state |0> data set, taking the qubit from the ground state |0> to the excited state |1> and then measuring the qubit state again using the level 1 measurement setup. A Gaussian pulse is applied.

This is repeated for the 1024 trails as well, generating a data set of excited states |1>, where the two known states in the entire data set become the actual labels and the in-phase and quadrature components become features.

Qiskit Pulse on IBM quantum computers provides access to quantum computation output in raw format.

Once the execution of the quantum algorithm is complete, a readout pulse is sent to the readout channel and the corresponding signal measuring the energy state of the qubit is recorded in the acquisition channel.

The signals recorded over the entire acquisition period are summed and a single complex value consisting of in-phase and quadrature (IQ) components is returned, as shown in graph 200.

The results in graph 200 are performed using measurement level 1 of Qiskit Pulse, and since the output is stochastic, multiple shots are required for a single experiment to obtain a good and reliable probability distribution.

The maximum number of shots allowed by IBM is 8192, and these multiple IQ values can be used to determine the state of the qubit.

Estimating the state of a qubit basically determines the probability amplitude value.

A machine learning-based classifier is used to create a decision boundary in the IQ plane and helps determine whether the qubit is in the ground state |0>, the excited state |1>, or the superposition state |ψ>.

Using Qiskit Pulse and the Python programming language, advanced and better classifiers can be designed on classical computers and used on quantum computers.

Therefore, the present invention can improve the performance of quantum state classification by evaluating the performance of various machine learning-based classifiers in hardware for qubit state estimation and selecting a classifier with excellent performance.

Additionally, the present invention can provide a Python-based platform design for computing, plotting, and comparing various machine learning base classifiers for qubit reading demonstrated on IBM quantum computers.

Figure 3 is a diagram illustrating the accuracy evaluation results by the classifier evaluation unit in the quantum state classification device according to an embodiment of the present invention.

Referring to FIG. 3, a graph 300 illustrates the accuracy evaluation results by the classifier evaluation unit in the quantum state classification device according to the type of machine learning.

Graph 300 illustrates the accuracy of classifiers based on LDA, SVM, KNN, and NNet.

There are several ways to judge the performance of a classifier, such as accuracy, recall, precision, F1 score, etc. Accuracy is used here.

The accuracy of all classifiers is calculated using the test data.

To obtain a better approximation of classifier performance, the training and testing process are each repeated 10 times with a new data set.

Values used include the median accuracy and the interquartile range (IQR) of accuracy (75th percentile - 25th percentile).

The accuracy of the classifier can be summarized in Table 1 below.

sorter Median interquartile range LDA 0.9594 0.0090 SVM 0.9594 0.0056 KNN 0.9541 0.0065 NNet 0.9580 0.0068

Classifiers using simple mathematical models such as LDA and SVM provide good performance with moderate accuracy and nearly linear decision boundaries.

In the case of KNN, it lagged in predicting datasets outside of the training data, creating an overfitted nonlinear decision boundary that gave it relatively poor performance.

In the case of NNet, randomly initializing the weights each time it trains provides mixed performance along with the fact that the feature set is not too complex for the neural network to explore its full potential.

Looking at the performance in terms of IQR of accuracy, SVM gives better results with smaller overall spread showing that SVM maintains its performance across multiple training and test traces on this particular type of in-phase and quadrature feature dataset.

This is important because IBM quantum computers are calibrated daily with their classifiers, so using SVMs as classifiers for qubit state estimation can help achieve better results on already noisy quantum computers, where even small improvements can help significantly. It can be helpful.

FIGS. 4A and 4B are diagrams illustrating an artificial neural network (ANN) model trained by a classifier training unit in a quantum state classification device according to an embodiment of the present invention.

Figure 4a illustrates linear optimization of the ANN model, and Figure 4b illustrates non-linear optimization of the ANN model.

In the ANN model 400, x ₀ represents a bias neuron, x ₁ to x ₄ represent input neurons, and y ₁ represents an output neuron.

In the ANN model 410, x ₀ represents a bias neuron, x ₁ to x ₄ represent input neurons, h ₁ to h _i represent hidden neurons, and y ₁ represents an output neuron.

The ANN model 400 is a linear optimization of weights without a hidden layer, and the ANN model 410 is a non-linear optimization of weights with a hidden layer introduced.

In order to train the Bell or CHSH (Clauser-Horne-Shimony-Holt) inequality as an ANN model-based classifier, which is a quantum state classifier, the quantum state can be separated as shown in Equation 1 below.

[Equation 1]

In Equation 1, p _sep can represent a separate quantum state, where p _i is greater than or equal to 0 and less than or equal to 1, and in Equation 1, p _i converges to 1. Otherwise the quantum states become entangled.

Regarding entanglement detection in partial information, it can be considered to generalize Bell's inequality into Equation 2 below.

[Equation 2]

In Equation 2, <.> represents the expectation and {a, a'} and {b, b'} can be the detector settings of the two involved parties, A (Alice) and B (Bob).

In the case of entangled states and separable states, the value of Bell's inequality can be used as partial information.

This inequality cannot be directly utilized as a reliable tool for entanglement detection due to the following problems.

A pure entangled state that does not violate the inequality can be defined as Equation 3.

[Equation 3]

In Equation 3, ψ can represent a quantum state, and then, as in Equation 4, a pure entangled state that is entangled but does not violate the inequality and other quantum states depending on the measurement angle are considered.

[Equation 4]

In equation (4), ψ can represent a quantum state, and for arbitrary values of (θ, Φ) the aforementioned entangled state no longer violates the inequality in a fixed metric.

Moreover, this criterion is not robust in noise, which limits its effectiveness for mixed-state entanglement detection.

In summary, the original CHSH inequality cannot be utilized as a reliable tool for the detection of binary entanglements.

To overcome the aforementioned limitations, we will use supervised machine learning techniques, namely artificial neural networks.

The original CHSH operator can be defined as Equation 5.

[Equation 5]

The variables in Equation 5 can be converted to the Pauli operator, where a ₀ is σ _z , a' ₀ is σ _x , and b ₀ is the value of σ _z subtracted from σ _x divided by √2. value, and b' ₀ may be the sum of σ _x and σ _z divided by √2.

This operator can include the vector (1, -1, 1, 1, -2) as a fixed weight vector and can be used as a feature vector to be used in training a machine learning model.

By applying weights to Equation 5, it can be expressed as Equation 6 below.

[Equation 6]

Expected values{ _{<a 0 b 0>} , _{<a 0 b'0>} , <a' ₀ b ₀ >, <a' ₀ b' ₀ >} are used as input features, and weights are randomly initialized and used throughout the learning process. Can be trained across

The labeling of the data is determined by the positive partial transpose (PPT) criterion, and finally, the optimized weights can be obtained by an artificial neural network that generalizes the CHSH entanglement detection range to an arbitrary state.

The accuracy of the classifier can be described as a function of the number of neurons in the hidden layer, and training data for the quantum state family can be derived as shown in Equation 7 below.

[Equation 7]

In Equation 7, p can represent additive noise, and |Ψ _θ,Φ > is defined through Equation 4 above.

Figure 4c is a diagram explaining the accuracy evaluation results for the ANN model according to an embodiment of the present invention.

Referring to FIG. 4C, graph 420 illustrates the accuracy of the ANN model-based classifier depending on the number of hidden layers in FIG. 4B.

Graph 420 shows that accuracy increases as the number of neurons in the hidden layer increases.

As accuracy does not improve after a certain number, an accuracy of about 91% can be secured considering a model with 20 neurons in the hidden layer.

Therefore, the present invention is an artificial neural network (ANN) with entanglement separability for arbitrary bipartite states based on traditional entanglement detection methods such as Bell or Clauser-Horne-Shimony-Holt (CHSH) inequality. It can be provided as a base classifier.

It can be provided to a trained ANN-based classifier that extracts partial information of the qubit requiring class estimation and outputs a label indicating the quantum state of the qubit.

The present invention provides quantum technology to reduce qubit readout errors, detect entanglement of qubits for efficient classification of states for quantum algorithms, and for applications in quantum computing, quantum error correction, and quantum machine learning. can be provided.

The present invention utilizes partial information of the quantum state instead of the entire density matrix, and enables the design of a classifier using the ANN model to easily detect entanglement for quantum state classification.

Figure 5 is a diagram explaining a quantum state classification method according to an embodiment of the present invention.

Figure 5 illustrates an embodiment in which the quantum state classification method according to an embodiment of the present invention trains a machine learning-based classifier model and tests the trained classifier model to classify the qubit state at the pulse level.

Referring to Figure 5, in step 501, the quantum state classification method according to an embodiment of the present invention generates a training and evaluation data set related to the qubit state.

In other words, the quantum state classification method estimates the ground state, excited state, and mixed state for the qubit state based on pulse level control, and the estimated ground state The ground state, excited state, and mixed state can be randomly shuffled to generate a training data set and an evaluation data set in a one-to-one ratio.

In step 502, the quantum state classification method according to an embodiment of the present invention trains a plurality of classifiers using a training data set.

In other words, the quantum state classification method uses a training data set to classify an LDA (linear discriminant analysis)-based classifier, SVM (support vector machine)-based classifier, KNN (k-nearest neighbor)-based classifier, NNet (neural network)-based classifier, and ANN. Learn multiple classifiers, including an (artificial neural network)-based classifier.

In step 503, the quantum state classification method according to an embodiment of the present invention evaluates a plurality of classifiers using an evaluation data set.

In other words, the quantum state classification method determines the performance of the classifiers by calculating the median and interquartile range in relation to the classification accuracy of classifying qubit read data at the pulse level for each of a plurality of classifiers, and compares the performances of the determined classifiers with each other. Thus, multiple classifiers can be evaluated.

In step 504, the quantum state classification method according to an embodiment of the present invention classifies the qubit state into an entangled state and a separable state using a classifier selected based on the evaluation.

That is, the quantum state classification method selects one of a plurality of classifiers based on the classification accuracy calculated in step 503, and uses the selected classifier to classify the qubit state into an entangled state or a separable state. It can be classified as (separatable state).

Figure 6 is a diagram for explaining a counterfactual entanglement distribution model according to an embodiment of the present invention.

Referring to FIG. 6, the first terminal (Alice) 610, the second terminal (Bob) 620, and the third terminal (Charlie) 630 can each operate as a transmitting terminal, a receiving terminal, and a control terminal. there is. For example, when Alice (610) sends quantum information to Bob (620), Charlie (630) acts as a control terminal, and conversely, it is also possible for Bob (620) to send quantum information to Alice (630). do. As such, the controlled quantum teleportation protocol according to an embodiment of the present invention assumes a three-party quantum network.

Meanwhile, in FIG. 6, reference numeral 640 'SW' refers to a quantum channel including a switchable mirror and a switchable rotor, QAO is an abbreviation for quantum absorptive object, and V-CQZ is a vertical ( Vertical) indicates that it contains a CQZ gate, and H-CQZ indicates that it contains a horizontal (Horizontal) CQZ gate.

At least one of the first terminal (Alice) 610, the second terminal (Bob) 620, and the third terminal (Charlie) 630 shown in FIG. 6 is the counterfactual quantum communication processor 2013 of FIG. 12. can be applied to

[Step 1]

In the first step, each terminal counterfactually generates 2N GHZ states.

To safely perform controlled counterfactual quantum teleportation, Alice, Bob, and Charlie use a CQZ (Counterfactual quantum Zeno) gate to counterfactually generate 2N GHZ states. Here, N represents a natural number, and each terminal generates a 2N tripartie GHZ state with one qubit of the 2N GHZ state.

Alice, assumed to be a transmitting terminal, has her qubit operate as a quantum absorptive object (QAO) in each round of GHZ state creation in order to counterfactually distribute entanglement. Prepare the qubit in this state. At this time, Bob and Charlie each and Prepare the qubit in the in state.

Bob and Charlie alternately connect with Alice through quantum channels using CQZ gates in each round of GHZ state generation, as shown in Figure 6.

At the end of each round, the initial states of Alice, Bob, and Charlie are converted as shown in Equation 8.

[Equation 8]

,

.

Here, A, B, and C are indices representing Alice, Bob, and Charlie, respectively, and i refers to the index representing the ith round of GHZ state creation.

[Second stage]

In the second step, Alice, Bob, and Charlie perform qubit measurements and announce the results.

First, for security check, Alice selects a random subset of N GHZ states among 2N GHZ states and announces the location of the selected subset.

Bob and Charlie randomly select a measurement basis from the {x, y} set, perform measurements on each qubit, and report the measurement results and the selected basis set.

Alice compares or matches the measurement results of Bob and Charlie by referring to the measurement result table after Bob and Charlie's measurements as shown in [Table 2]. Through this, you can check whether the measurement results are consistent.

[Table 2]

If the consistency of the measurement results is reliable up to a preset acceptable threshold, Alice, Bob, and Charlie can perform the next steps. That is, Alice, Bob, and Charlie can transmit arbitrary unknown qubits in a tripartite quantum network using N GHZ states other than the subset selected in the next step.

Meanwhile, considering that each party can operate as a transmitting terminal, receiving terminal, and control terminal, the GHZ state subscripts will be expressed as s, r, and c in the following description. Here, s, r, and c represent qubits provided (or owned by the corresponding terminal) in the transmitting terminal, receiving terminal, and control terminal, respectively.

[Step 3]

In the third step, the transmitting terminal teleports the unknown quantum state by performing a bell-based measurement on the qubit it owns. More specifically, the unknown quantum state defined as Equation 9: Assume that the transmitting terminal transmits.

[Equation 9]

,

.

The combination state of the unknown quantum state and the GHZ state can be expressed as Equation 10.

[Equation 10]

,

.

here, and represents the bell state, and j represents the jth round of controlled quantum teleportation.

In the third step, the transmitting terminal performs Bell basis measurements on qubits q and s _j , and records and maintains the measurement results.

[Step 4]

In the fourth step, the control terminal measures the qubit based on the Hadamard basis and notifies the receiving terminal of the measurement result to reconstruct the unknown quantum state.

More specifically, the control terminal performs the Hadamard gate H on qubit c _j and measures the qubit as a computational basis. The control terminal notifies the measurement result through 1 bit of classical information m to approve the controlled counterfactual quantum teleportation.

When the control terminal approves communication between the sending terminal and the receiving terminal, the sending terminal is in the ring state. corresponding to The measurement result is notified to the control terminal and the receiving terminal through 2 bits of m1m2, which is classical information from the set.

Figure 7 is a diagram showing the concept of controlled quantum teleportation according to an embodiment of the present invention.

Referring to FIG. 7, it can be seen that the transmitting terminal performs controlled quantum teleportation by notifying the bell measurement result with the approval of the control terminal.

[Step 5]

In the fifth step, the receiving terminal reconstructs the quantum state.

More specifically, the receiving terminal performs a unitary operation defined as Equation 11 on the qubit r _j based on the measurement results of the transmitting terminal and the control terminal to reconstruct the quantum state.

[Equation 11]

Here, X and Z refer to the Pauli x and z operators, respectively.

The method described through steps 1 to 5 can reconstruct an unknown quantum state sequence under the approval of a control terminal performing control. Therefore, it is safer against various attack strategies and can be used in a multi-party communication system.

If a terminal attempting eavesdropping attempts to participate in the entanglement distribution process, it may be in an entanglement relationship, but the terminal attempting eavesdropping may be detected during the security check process of the second stage.

Additionally, if a terminal attempting eavesdropping adds invisible photons to steal Alice's unknown quantum state, more errors may occur in the security check step due to the invisible photons.

Figure 8 is a flowchart showing the operation of a controlled counterfactual quantum teleportation support device according to an embodiment of the present invention.

Here, the controlled counterfactual quantum teleportation support device may be any one of a transmitting terminal, a receiving terminal, and a control terminal, and Figure 8 corresponds to the case where the controlled counterfactual quantum teleportation support device is a transmitting terminal.

Referring to FIG. 8, the transmitting terminal counterfactually generates 2N GHZ (Greenberger-Horne-Zeilinger) states in step 810.

In step 820, the transmitting terminal selects a random subset of N GHZ states and notifies the location of the selected subset.

In step 830, the transmitting terminal checks whether the measurement results are consistent.

If the consistency of the measurement result is reliable up to a preset acceptable threshold, the transmitting terminal performs bell-based measurement on the owned qubit in step 840.

Thereafter, if communication is approved by the control terminal, in step 850, the transmitting terminal notifies the measurement result using the classic 2 bits corresponding to the bell state.

Hereinafter, a quantum state tomography processing unit that can be applied to the quantum experiment platform in the quantum information technology integrated system of the present invention will be described.

Quantum State Tomography (QST) is a quantum information processing task to determine unknown quantum states. This work is performed by measuring an ensemble of identically prepared quantum systems and inferring their states using various estimation methods. When the quantum state passes through the maximum noise depolarization channel with a clear causal sequence (classical trajectory) input before measurement, the quantum state switches to the maximum mixed state, making QST impossible.

Instead of propagating quantum states through classical trajectories, the proposed method here sets the channels in an infinite causal order (ICO). In this specification, we show that by setting a depolarizing channel in ICO, QST can be performed even if the channel is maximally noisy. And/or, this specification examines the impact of ICO on QST through linear regression estimation when the channel has no maximum noise. And/or, the present disclosure shows that configuring channels in an ICO arrangement is advantageous when the channels have maximum noise.

The quantum state tomography processing unit according to an embodiment of the present invention transmits the quantum state to be estimated to the quantum switch, performs quantum state tomography in the state output from the quantum switch, and calculates the output of the quantum state tomography and algebra. Based on this, the initial state can be estimated.

Quantum work that determines the state of an unknown system can be called quantum state tomography (QST). QST can consist of two steps.

First, measurements are performed on identically prepared copies of the quantum state.

Next, the quantum state is reconstructed using classical statistical estimation methods. Methods to construct a quantum state are linear inversion (MLE), maximum likelihood estimation, Bayesian estimation method, and linear regression. There may be.

Linear inversion can be performed assuming an ideal environment. Therefore, if an error occurs during qubit measurement, the recovered state may not be the physical state. MLE finds the density matrix that provides the highest probability of generating a set of measurement data determined by numerical optimization. The Bayesian method uses the Bayesian theorem to determine the posterior probability distribution p(ρ|D) with knowledge of the prior probability distribution p(ρ). Here, D means measurement data.

In traditional quantum communication, quantum messages are a series of communication channels. and When transmitted through a channel, the order of messages transmitted through the channel is After passing or first After passing through It can be defined as passing through.

Quantum theory could allow new ways to combine quantum channels by creating superpositions of these channels. Through this, the information transfer order does not correspond to the previous two states. That is, the causal order between quantum channels is infinite. The process of infinite causal order can be achieved by a quantum switch where the order of operations is controlled by quantum bits. And/or, information can be transmitted through a combination of fully depolarizing channels, superimposing two alternative sequences. Quantum switches can be used to improve quantum metrology, enhance channel capacity, identify quantum depolarization channels, and enhance the quantum internet.

Herein, we model QST through linear regression under the assumption that the quantum state is transmitted through a quantum depolarization channel with maximum noise before measurement. Additionally, in this specification, we examine the impact of setting the quantum channel to an infinite causal order on QST. And, in this specification, the infidelity of the estimated state and the actual state is compared when transmitted through a channel assembled with a definite causal sequence and ICO. The latter method can support QST for quantum depolarization channels with maximum noise.

Hereinafter, in this specification, we first look at quantum state tomography using a linear regression method and a quantum switch, and then look at QST performance when transmitting quantum states through a channel established according to a clear causal order and ICO.

Quantum State Tomography

First, we look at a linear regression approach to estimate unknown quantum states.

Kronecker's delta function, equal to 1 if can be used. means tracking the matrix (1) and (2) Hermitian operator that satisfies It can be specified as . For example, in the single-qubit case, may be the same as Equation 12 below.

[Equation 12]

The qubit density operator can be parameterized as shown in Equation 13 below.

[Equation 13]

here, and the measurement set These basis It can be assumed that it can be parameterized as shown in Equation 14 below.

[Equation 14]

here, ego, am. The measurement set used in this work is This is the Stokes measurement set. Based on Born's rule, if multiple copies of an identically prepared quantum system are measured, The probability of obtaining the measurement result may be as follows in Equation 15.

[Equation 15]

here, is a column vector is the transpose of am.

From Equation 15 above, the linear regression form for QST may be as shown in Equation 16 below.

[Equation 16]

here, ego, am.

each About status It can be assumed that the quantum system is measured N times. Measurement results The probability of obtaining can be estimated by the following equation 17.

[Equation 17]

here, is the estimation error represents the observed frequencies with . Next, the linear regression equation in Equation 16 may be as shown in Equation 18 below.

[Equation 18]

here, ego, am. For large N, is the mean 0 and the change It can converge to a distribution for a normal distribution with . If more measurements are performed, measurements of the qubit state can become more accurate. The solution to the linear regression problem in Equation 18 above is to minimize Equation 19: It can be.

[Equation 19]

measurement set This information is complete or overflowing. It can be assumed that it has a station. estimate is an unbiased estimate with a mean squared error, and asymptotically may be equivalent to Equation 20 below.

[Equation 20]

here, am.

Quantum Depolarizing Channel

Quantum information can be transmitted through quantum channels. A quantum channel can be mathematically defined as a complete positive trace that preserves the map over the density operator. While transmitted through the channel, the qubit can evolve into a fully mixed state with probability (1-q) and remain with probability q. This type of quantum channel is called quantum depolarization channel It can be called . Evolution of qubits is defined as in Equation 21 below.

[Equation 21]

Quantum Switch

As described above, a quantum system transmitted through two quantum channels evolves in a clear causal sequence, with the second channel acting on the output of the first channel. Because quantum systems can evolve even in ICOs, it is impossible to determine which process occurs before the other. This is possible because of quantum switches. Quantum switches combine quantum channels, qubits called control qubits. You can control the relative order by . control qubit When in the state, the quantum switch is We will go through quantum systems in order. control qubit If in a state, the quantum state is It passes through quantum channels in order. control qubit together with and If there is a superposition of , the quantum switch changes the channel to and It can be placed in a nested order. As a result of the quantum switch, the Kraus operator of the quantum channel can be defined as Equation 22 below.

[Equation 22]

Here, channel and The Kraus operators are respectively and It can be expressed as

The quantum switch for one qubit system can be given as Equation 23 below.

[Equation 23]

Tomography methods, simulations and results

Quantum state to infer It can be assumed that is transmitted to the measurement system through a noisy depolarizing channel in the classical trajectory. If the channel is at maximum noise, e.g. When , from Equation 10, the qubit is It can be converted to the maximum mixed state of Equation 24 below, which does not depend on .

[Equation 24]

as a result, If QST is performed in It is impossible to reconstruct.

control qubit (here, ), if the channel is set up in an infinite causal sequence, the output of the channel can be as shown in Equation 25 below.

[Equation 25]

When measuring the control qubit in basis, the output of the channel can be converted as shown in Equation 26 below.

[Equation 26]

here, and are respectively and can represent the probability of obtaining a measurement result. If the channel is maximally noisy, from equation 26: If you perform QST on is the state meaning that it is possible to reconstruct can depend on

If you measure not You will get an estimate of

Whether the probability of a quantum state remains intact is Can be determined by quantum process tomography from Equation 26, which can be recovered as shown in Equation 27 below.

[Equation 27]

If Equation 27 is applied to the above, There may be some cases where is a non-physical matrix.

For this reason, can be projected onto the physical matrix space and expressed as a space of unit traces and positive semidefinite matrices. A simple projection algorithm can be adopted, the idea of which may be to transform all the eigenvalues of the density matrix into positive values and their sum.

<MSE analysis>

If the quantum state in Equation 26 is measured, then the state for The probability of obtaining the measurement result of can be defined as Equation 28 below.

[Equation 28]

situation In the case of , the certainty can be defined as follows in Equation 29.

[Equation 29]

For the maximum noise channel (q=0), may be equal to the following equation 30:

[Equation 30]

Here, the control qubit is in the state In the case that applies, that is, In this case, better results can be obtained.

From equation 20, X can obtain the same value for both states obtained after measuring the control qubit. Therefore, there is a need to identify the differences caused by P.

To assume a contradiction, Can be expressed as Equation 33 through the following process from Equation 31 to Equation 32.

[Equation 31]

[Equation 32]

[Equation 33]

here, ego, and ego, and am.

Here, equation 33 is It can be expressed as is a contradiction. therefore, It can be done.

From Equation 24, if the quantum state is transmitted through a classical arrangement, The probability of obtaining a measurement result of And, accordingly, and A larger MSE will be given.

<Numerical Simulation>

Below, numerical simulations are performed on a single qubit system using the above-described scheme. QST is performed in pure and mixed states to ensure that the system performs well in both states.

FIG. 9 is a diagram illustrating a mechanism for performing QST in a quantum depolarization channel with maximum noise by employing a quantum switch 1010 in qubit transmission.

Referring to FIG. 9, the tomography device 1000 according to an embodiment configured during simulation includes a quantum switch 1010, a control qubit measurement device 1020, and , may include a quantum state tomography device 1030.

uniformly distributed according to the Haar measure and according to the Hilbert-Schmidt measure or Bures measure for mixed states. Random quantum states can be created. It is possible to model qubit transport in depolarizing channels and ICOs with maximum noise assembled from classical orbits prior to measurement. Measurements are made on each qubit. It can be performed every time. To quantify how close the actual state is to the estimate, both states The fidelity of all can be calculated.

next, Average log accuracy of quantum states can be calculated. If the states are transmitted in finite causal order and ICO, the Log infidelity of the actual pure states with this estimate can be represented as in Figure 10.

Figure 10 shows the reconstruction states for pure states (a) and mixed states (b) and actual state Indicates the log infidelity.

When real states are transmitted over channels placed in an ICO array, the log infidelity is and Measurement results can be displayed by red and green lines. Additionally, the average accuracy of the estimated state can be calculated as shown in Equation 34 below.

[Equation 34]

here, and is the measurement of the control qubit and In the case corresponding to , it may be the probability and accuracy of the estimated state.

From Figure 10, it can be seen that for the quantum depolarization channel with maximum noise, the ICO arrangement helps the QST estimation performance for both pure and mixed states. Measurement results The QST on the corresponding states is consistent with the previously described MSE analysis. A better estimate can be obtained than the corresponding states.

For example, the tomographic imaging device 1100 according to the embodiment of FIG. 11 may perform the above-described operation. For example, the output passing through the quantum switch unit 1110 of the tomography apparatus 1100 may be expressed as Equation 25. And/or, the quantum switch unit 1110 of the tomography device 1100 The output of the channel from the control qubit in the basis can be converted as shown in Equation 26. And/or, the tomography unit 1120 and/or the estimation unit 1130 of the tomography apparatus 1100 may estimate the quantum state based on a linear regression approach, as described above for quantum state tomography. . For example, the tomography unit 1120 and/or the estimation unit 1130 of the tomography apparatus 1100 may perform the calculations of Equations 12 to 20. And/or, the tomography unit 1120 and/or the estimation unit 1130 of the tomography apparatus 1100 based on Equation 26 can also be reconstructed.

And/or, the tomography unit 1120 and/or the estimation unit 1130 of the tomography apparatus 1100 is calculated by Equation 23: to or can be determined and/or estimated.

The operation of the tomography apparatus described with reference to FIG. 11 is not limited to the above-described method, and the operations described herein can be performed in various ways. Additionally, the tomographic imaging device may perform all or some of the operations described in this specification in addition to the operations of the tomographic imaging device described with reference to FIG. 11 above.

The tomography apparatus of FIG. 11 can be applied to the quantum state tomography processing unit 2031 of FIG. 12.

In this specification, we look at QST in a depolarizing channel with maximum noise. In this specification, a quantum switch was applied to enable QST in the maximum noise channel. As described herein, transmitting qubits in a quantum switching array provides more accurate quantum state estimates than transmitting them via classical orbitals. In other words, the present invention can perform accurate quantum state tomography in a depolarized quantum channel with maximum silencing.

Figure 11 is a block diagram of a tomographic imaging device according to an embodiment of the present invention.

Referring to FIG. 11, the tomography apparatus 1100 according to an embodiment of the present invention includes a quantum switch unit 1110 that is controlled by a control qubit and receives a quantum state to be estimated, and a quantum switch unit 1110. It includes a tomography unit 1120 that performs quantum state tomography on the output state, and an estimation unit 1130 that estimates the initial state based on the output of the tomography unit 1120 and algebra, even in the maximum noise channel. Quantum state imaging can be performed.

And/or, the quantum switch unit 1110 may combine quantum channels.

And/or, tomography may be based on linear regression estimation.

And/or, the maximum noise channel can be set to Indefinite Causal Order (ICO).

And/or, the maximum noise channel may be a quantum depolarization channel.

According to another embodiment, the tomography device 1100 transmits the quantum state to be estimated to the quantum switch, performs quantum state tomography in the state output from the quantum switch, and uses algebra in predicting the state output from the quantum switch. By performing , the quantum state tomography accuracy can be improved in the maximum noise depolarization quantum channel by estimating the initial state.

According to another embodiment, the tomography device 1100 transmits the quantum state to be estimated to the quantum switch, performs quantum state tomography in the state output from the quantum switch, and uses algebra in predicting the state output from the quantum switch. By performing to estimate the initial state, quantum state tomography can be performed even when the channel noise is maximum.

And/or, when the quantum state is transmitted in a maximally noisy depolarizing quantum channel, it transforms into a maximally mixed state, making it impossible to perform quantum state tomography. Even when the channel noise is maximum, the quantum state tomography device according to one embodiment first transmits the quantum state to be estimated to the quantum switch, performs quantum state tomography in the state output from the quantum switch, and outputs from the quantum switch. The initial state can be estimated by performing algebra on the prediction of the resulting state.

The operations of the tomography apparatus described with reference to FIG. 11 are not limited to this, and all or some of the operations described in this specification may be performed by the tomography apparatus.

Figure 12 is a diagram showing a quantum information technology integrated system according to an embodiment of the present invention.

Referring to FIG. 12, it includes a quantum communication platform (2010), a quantum computing platform (2020), a quantum observation and analysis platform (2030), a platform control unit (2040), and a platform interface unit (2050).

Quantum Communication Platform (2010) is a platform that provides implementation and experimentation of element technologies related to quantum communication and quantum information communication.

Quantum communication platform (2010) is a quantum key distribution processor (2011) that performs secure communication through quantum key distribution (QKD), particle-free information transmission in which no particles exist in the channel. and a counterfactual quantum communication processing unit (2013) that performs a new form of full-duplex information transmission and blind tele-computation, and radio resource management when applying quantum communication to next-generation mobile communication. It may include a radio resource management processing unit 2015 that performs and a network interference control processing unit 2017 that performs network interference control. For example, the counterfactual quantum communication processing unit 2013 may perform the role of at least one of the first terminal (Alice), the second terminal (Bob), and the third terminal (Charlie).

The Quantum Computing Platform (2020) is a platform that provides implementation and experiments of element technologies for the practical use of quantum computers capable of exponentially fast parallel calculations based on quantum algorithms.

The quantum computing platform (2020) includes a variant quantum algorithm processor (2021) that provides a variational quantum algorithm (VQA) specifically designed for noisy intermediate-scale quantum (NISQ) devices, the Grover search algorithm or Quantum search algorithm processing unit (2023) that provides a quantum search algorithm (QSA) such as a quantum counting algorithm and quantum machine learning such as quantum reinforcement learning of a multi-body Hamiltonian It may include a quantum machine learning processing unit (2025) that provides learning: QML) technology.

The Quantum Observation and Analysis Platform (2030) supports quantum measurements and quantum simulations and directly utilizes the quantum properties of real particles, such as quantum coherence and entanglement, solving problems that are difficult to simulate on classical computers. It is a platform for doing this.

The quantum observation and analysis platform (2030) includes a quantum state tomography processor (2031) that provides quantum state tomography, a quantum state measurement processor (2033) that supports quantum state measurements, and quantum correlation measurements for qubit measurements and experiments. It includes a quantum correlation measurement processing unit 2035 and a tensor network experiment processing unit 2037 supporting tensor network and Hamiltonian-based quantum simulation for simulation.

The platform control unit 240 stores the history and measurement results performed on the quantum communication platform (2010), the quantum computing platform (2020), and the quantum observation and analysis platform (2030), and controls the components of the other platform from any one platform. When you need to use it, you can perform an action to connect it.

For example, if a Bell measurement is required in the counterfactual quantum communication processing unit (2013) of the quantum communication platform (2010), the Bell measurement is performed in the quantum state measurement processing unit (2033) or the quantum correlation measurement processing unit (2033) of the quantum observation and analysis platform (2030). 2035).

The platform interface unit 2050 may be a computer device that allows a user or experimenter to access each platform. For example, a user can access the quantum observation and analysis platform 2030 through the platform interface unit 2050 and run a desired quantum simulation.

For example, the quantum information technology integrated system according to an embodiment of the present invention performs the operations described in Figures 9 to 11 and a quantum communication platform including a counterfactual quantum communication processing unit that performs the first to fifth steps. It may include a quantum observation and analysis platform including a quantum state tomography processing unit.

Quantum information technology is converging with the fields of communication, algorithms, and simulation technologies, and research is ongoing to overcome the limitations of next-generation mobile communications, such as network security, wireless resource management, channel estimation, and network interference management. Accordingly, the components included in each of the quantum communication platform (2010), the quantum computing platform (2020), and the quantum observation and analysis platform (2030) may be replaced or improved according to the development and emergence of each technology.

For example, the quantum key distribution processor 2011 can be improved to provide technology for integration into the security layer of the next-generation mobile communication system.

For example, each component of the quantum communication platform (2010) can be improved for use in interference management of next-generation mobile communication systems that require intelligent network functions.

Although the embodiments have been described with limited drawings as described above, various modifications and variations can be made by those skilled in the art from the above description. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, area, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent.

Therefore, other implementations, other embodiments, and equivalents of the claims also fall within the scope of the claims described below.

100: Quantum state classification device
110: Data set generation unit 120: Classifier training unit
130: Classifier evaluation unit 140: Quantum state classification unit

Claims (12)

A data set generator that generates a training data set and an evaluation data set related to the qubit state;
a classifier training unit that trains a plurality of classifiers using the generated training data set;
Calculate the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the generated evaluation data set, and evaluate the plurality of classifiers based on the calculated classification accuracy. a classifier evaluation unit; and
Characterized by comprising a quantum state classification unit that classifies the qubit state into an entangled state and a separable state using any one of the plurality of classifiers based on the calculated classification accuracy. doing
Quantum state classification device. According to paragraph 1,
The data set generator estimates a ground state, an excited state, and a mixed state for the qubit state based on pulse level control, and the estimated ground state Characterized in generating the training data set and the evaluation data set in a one-to-one ratio by randomly shuffling the ground state, excited state, and mixed state.
Quantum state classification device. According to paragraph 1,
The plurality of classifiers include a linear discriminant analysis (LDA)-based classifier, a support vector machine (SVM)-based classifier, a k-nearest neighbor (KNN)-based classifier, a neural network (NNet)-based classifier, and an artificial neural network (ANN)-based classifier. Characterized by including
Quantum state classification device. According to paragraph 3,
The classifier training unit trains the ANN (artificial neural network)-based classifier using Bell or CHSH (Clauser-Horne-Shimony-Holt) inequality values as feature vectors of the generated training data set. characterized by
Quantum state classification device. According to clause 4,
The feature vector is partial information of the qubit and is calculated as a weight by an artificial neural network, and the range of entangled state detection of the Bell operator or CHSH operator is generalized to an arbitrary state.
Quantum state classification device. According to clause 5,
The artificial neural network is constructed using either linear optimization or non-linear optimization, and when constructed using non-linear optimization, classification accuracy increases as the number of neurons in the hidden layer increases.
Quantum state classification device. According to paragraph 1,
The classifier evaluation unit determines the performance of the classifier by calculating the median and the interquartile range in relation to the classification accuracy for classifying qubit read data at the pulse level for each of the plurality of classifiers, and determines the performance of the determined classifier. Characterized by evaluating the plurality of classifiers by comparing them with each other
Quantum state classification device. Generating, in a data set generator, a training data set and an evaluation data set related to qubit states;
In a classifier training unit, training a plurality of classifiers using the generated training data set;
In the classifier evaluation unit, the median and interquartile range of classification accuracy for the plurality of trained classifiers are calculated based on the generated evaluation data set, and the median and interquartile range are calculated based on the calculated classification accuracy. evaluating a plurality of classifiers; and
In a quantum state classification unit, classifying the qubit state into an entangled state and a separable state using any one of the plurality of classifiers based on the calculated classification accuracy. characterized by
Quantum state classification method. According to clause 8,
The step of generating a training data set and an evaluation data set related to the qubit state,
Based on pulse level control, the ground state, excited state, and mixed state are estimated for the qubit state, and the estimated ground state, Characterized by generating the training data set and the evaluation data set in a one-to-one ratio by randomly shuffling excited states and mixed states.
Quantum state classification method. According to clause 8,
The plurality of classifiers include a linear discriminant analysis (LDA)-based classifier, a support vector machine (SVM)-based classifier, a k-nearest neighbor (KNN)-based classifier, a neural network (NNet)-based classifier, and an artificial neural network (ANN)-based classifier. Characterized by including
Quantum state classification method. According to clause 10,
The step of training a plurality of classifiers using the generated training data set is,
Comprising the step of training the ANN (artificial neural network)-based classifier using Bell or CHSH (Clauser-Horne-Shimony-Holt) inequality values as feature vectors of the generated training data set. characterized by
Quantum state classification method. According to clause 8,
Calculate the median and interquartile range of classification accuracy for the plurality of trained classifiers based on the generated evaluation data set, and evaluate the plurality of classifiers based on the calculated classification accuracy. The steps are:
For each of the plurality of classifiers, the median and the interquartile range are calculated in relation to the classification accuracy for classifying qubit read data at the pulse level to determine the performance of the classifiers, and the performances of the determined classifiers are compared with each other to determine the median and the interquartile range. Characterized by comprising the step of evaluating a plurality of classifiers
Quantum state classification method.