Method and system for resisting quantum noise interference based on quantum random dethronement

Abstract

The present invention belongs to the technical field of resisting quantum noise interference, and discloses a method and system for resisting quantum noise interference based on quantum random dethronement. The present invention designs a new experimental environment, i.e., using a classical computer to simulate a noisy quantum circuit, in order to solve the problems that the previous quantum-classical hybrid neural network lacks consideration of quantum noise and the large-scale quantum circuit, such as quantum convolutional neural network, seriously decreases in the accuracy of image recognition and classification under the interference of quantum noise. This makes the quantum neural network model in the classical computer environment basically consistent with the experimental results in the real quantum device environment, and the experimental results are more practical. Secondly, the algorithm running on the quantum computer can also be simulated on the classical computer, which greatly reduces the cost of the experiment.

Description

Technical Field

[0001] The present invention belongs to the technical field of resisting quantum noise interference, and in particular relates to a method and system for resisting quantum noise interference based on quantum random dethronement. Background Art

[0002] Quantum-classical hybrid machine learning, which combines quantum computing with classical computers, has become the main way to implement quantum machine learning. This method combines the advantages of quantum computers, such as superposition and entanglement, with the stability and ease of use of classical computers, providing new ideas and tools for solving complex problems. Quantum computers gain powerful computing power through the parallelism between quantum bits, but due to the limitations of the hardware level of experimental equipment, quantum circuits are not pure. When conducting quantum circuit experiments, quantum noise such as decoherence, phase flip, and depolarization are often encountered. These noises will affect the calculation results and limit the accuracy and stability of quantum computing.

[0003] In previous quantum-classical hybrid machine learning experiments, an important influencing factor is often overlooked - quantum noise. Since quantum computing has attracted widespread interest in academia, the errors caused by quantum noise in experimental results have raised questions about the feasibility of quantum machine learning. Recently, some researchers have successfully achieved quantum error correction on quantum circuits through quantum simulations around short-depth quantum circuits and approximate optimization algorithms (QAOA). However, in the field of image recognition, how to implement quantum error correction in its representative quantum convolutional neural network (QCNN) circuits has not yet been solved. Due to the existence of quantum noise, QCNN may have overfitting problems in practical applications, resulting in classification accuracy far below the ideal situation.

[0004] Through the above analysis, the problems and defects of the prior art are as follows:

[0005] (1) There is a lack of simulation of quantum noise in actual quantum-classical hybrid circuits. That is, previous quantum-classical hybrid circuits were simulations under ideal conditions, and their experimental results were not general.

[0006] (2) Due to the high cost of quantum computers, the verification costs of many quantum machine learning experiments are high.

[0007] (3) In a quantum noise environment, the efficiency of many quantum machine learning algorithms will be greatly affected. How to achieve quantum error correction in a ubiquitous quantum noise environment is a difficult problem.

[0008] (4) At present, there is no effective method for how quantum convolutional neural networks can achieve quantum error correction in a quantum noise environment. A large amount of research is still at the theoretical level and on small-scale networks, and the problems that can be solved are also very limited. Summary of the invention

[0009] In view of the problems existing in the prior art, the present invention proposes a method and system for resisting quantum noise interference based on quantum random dethronement.

[0010] The present invention adds a quantum noise module on top of the existing ideal quantum convolutional neural network to construct a quantum convolutional neural network containing quantum noise; and adds an anti-quantum noise method based on quantum random dethronement to this quantum circuit to resist the adverse effects of noise on image recognition accuracy.

[0011] A method for resisting quantum noise interference based on quantum random dethronement includes:

[0012] Step 1: Data collection: download the MNIST dataset and Fashion-MNIST dataset;

[0013] Step 2: Extract the required training samples from the MNIST dataset and Fashion-MNIST dataset and shuffle the sample order;

[0014] Step 3: preprocess the data, including image downsampling and normalization;

[0015] Step 4: Build a quantum noise model and deploy it on a quantum convolutional neural network.

[0016] Step 5: Establish a quantum convolutional neural network model containing quantum noise and train the data set;

[0017] Step 6: Build a quantum noise-resistant VQC and deploy it on a noisy quantum convolutional neural network;

[0018] Step 7: Train the quantum noise-resistant model by adjusting different noise intensities;

[0019] Step 8: By comparing the classification accuracy and loss curve evaluation model before and after the addition of the quantum noise-resistant model, the quantum noise-resistant VQC and traditional deep learning algorithms are combined to achieve noise resistance in quantum circuits.

[0020] Furthermore, in step 4, the established quantum noise model adds noise to the ideal quantum circuit so that the experimental environment is more in line with the experimental environment of an actual quantum computer, including:

[0021] (1) Use Bitflip, Phaseflip, and Depolarizing Channel methods to simulate the situation when quantum bits are disturbed by quantum noise;

[0022] (2) The quantum data affected by quantum noise is transformed into classical data through the subsequent quantum measurement module.

[0023] Furthermore, the quantum convolutional neural network model containing quantum noise in step five includes the following seven groups of modules: data preprocessing module, quantum coding module, quantum entanglement module, quantum noise module, quantum measurement module, classification module, and data post-processing module.

[0024] Furthermore, the function of the data preprocessing module is to use dimensionality reduction, down sampling, and data flattening methods to process the input data to meet quantum requirements; the function of the quantum coding module is to use an angle coding method, and each channel is encoded with a quantum bit to realize the conversion of classical data to quantum state; the quantum entanglement module is to form a quantum convolution layer through Hadamard gates, single-qubit rotation gates and CNOT gates, and perform unitary transformation on the encoded quantum data to realize the transformation and entanglement of the quantum state; the function of the quantum measurement module is to measure the transformed and entangled quantum state to obtain the expected value with Hamiltonian, and pass the result to the classical post-processing module.

[0025] Furthermore, the function of the data post-processing module is to perform gradient training on the obtained expectations and the defined loss function, and dynamically update the entanglement module training parameters.

[0026] Furthermore, the quantum noise-resistant VQC in step 6 is inspired by the traditional deep learning idea. It reduces the expressibility of quantum convolutional neural network by randomly removing quantum gates in the circuit using a certain strategy. The strategy includes:

[0027] ① Randomly remove the CNOT gate or Pauli-Z gate containing entanglement in the quantum circuit during the i-th parameter update;

[0028] ② Calculate the gradient of the loss function generated after removing the entangled quantum gate, determine the i+1th parameter by the gradient descent method, and update the parameter;

[0029] ③ Repeat the above steps until the accuracy of the model reaches the expected value.

[0030] Another object of the present invention is to provide a quantum noise interference resistance system based on quantum random dethronement, comprising:

[0031] ① Data acquisition module, used to download the MNIST dataset and Fashion-MNIST dataset from the Internet and import them into the experimental platform;

[0032] ②Data preprocessing module, used to cut and reduce the dimension of data set images to meet the data input requirements of the quantum platform;

[0033] ③Data encoding module, which is used to encode classical data into quantum data, using quantum angle encoding. The encoded data participates in subsequent calculations and training;

[0034] ④ Quantum entanglement module, used to entangle quantum data, realize quantum data entanglement in the constructed quantum convolutional neural network, and train image classification tasks;

[0035] ⑤Quantum noise module, which is used to establish a quantum noise model and deploy it on a quantum convolutional neural network to realize the experimental environment of a noisy quantum neural network;

[0036] ⑥Quantum measurement module, used to measure quantum data and convert it into classical data, so as to facilitate training and updating of the subsequent data post-processing module;

[0037] ⑦ Quantum post-processing module, which is used to train and optimize the classical data after measurement, and uses the gradient descent method to train to achieve the desired effect;

[0038] ⑧Classification module, used to classify classic data and achieve the expected image classification task;

[0039] ⑨Anti-quantum noise module is used to combat the adverse effects of quantum noise on the classifier. The model is evaluated by two indicators: classification accuracy and loss function. The out-of-sample error is used to analyze the effects of different hyperparameters on the anti-noise model.

[0040] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for resisting quantum noise interference based on quantum random demotion.

[0041] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the method for resisting quantum noise interference based on quantum random demotion.

[0042] Another object of the present invention is to provide an information data processing terminal, which is used to implement the quantum random demotion-based anti-quantum noise interference system.

[0043] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0044] First, the significance of solving the problems and defects of the existing technology is that the quantum noise-resistant algorithm that originally required quantum computer experiments can also be simulated on classical computers, thereby saving a lot of costs; the proposed quantum noise-resistant algorithm is deployed in a large-scale quantum neural network, and this method successfully reduces the adverse effects of the quantum noise that is prevalent in the training process on classification accuracy; this marks a major progress in the application of quantum machine learning in actual production.

[0045] In view of the fact that previous quantum-classical hybrid neural networks lacked consideration of quantum noise and that large-scale quantum circuits, such as quantum convolutional neural networks, seriously decreased in accuracy in image recognition and classification under the interference of quantum noise, the present invention designed a new experimental environment, namely, using classical computers to simulate noisy quantum circuits. This makes the quantum neural network model in a classical computer environment basically consistent with the experimental results in a real quantum device environment, and the experimental results are more practical. Secondly, the algorithm running on a quantum computer can also be simulated on a classical computer, which greatly reduces the cost of the experiment.

[0046] The present invention first performs quantum encoding on the data set in a noise-free quantum environment, and then performs three-classification of the results. Under ideal conditions, the accuracy of the three-classification exceeds 90%. After that, a noise model is added to simulate the classification in a quantum noise environment, and the classification accuracy drops sharply to 50%-60%. After adding the anti-noise model, the classification accuracy rises to more than 85%, close to the level in an ideal environment, indicating that the invention has played a good effect in resisting noise interference in large-scale quantum networks.

[0047] Second, the embodiment of the present invention provides a method for resisting quantum noise interference based on quantum random dethronement, and proposes an effective solution to the noise problem faced by quantum computing when processing complex data tasks. The following are the technical problems solved by the present invention and the significant technical progress achieved:

[0048] Technical issues solved:

[0049] 1. Noise interference in quantum computing:

[0050] In quantum computing, noise is the main cause of quantum information loss and quantum algorithm execution errors. Existing quantum computing models have not yet fully addressed how to effectively resist the noise interference inherent in quantum circuits.

[0051] 2. The robustness of quantum algorithms is insufficient:

[0052] Although quantum computing can theoretically provide performance that surpasses traditional computing, quantum algorithms in practical applications are often very sensitive to noise, which limits their scope and effectiveness in practical problems.

[0053] 3. Stability of quantum machine learning models:

[0054] When quantum machine learning is faced with actual data sets, such as MNIST or Fashion-MNIST, the presence of noise greatly reduces the learning ability and classification accuracy of the model, and it is necessary to enhance its stability and accuracy in noisy environments.

[0055] Significant technological advances achieved:

[0056] 1. Improved the robustness of quantum models:

[0057] By introducing a quantum anti-noise model and deploying it in a quantum convolutional neural network, the present invention significantly improves the robustness of the model in a noisy environment. This enables the quantum machine learning model to maintain high performance in a practical noisy environment and expands the application scope of quantum algorithms.

[0058] 2. Optimized noise management of quantum networks:

[0059] By constructing a quantum noise model and adjusting different noise intensities to train an anti-noise model, the present invention effectively simulates and manages the noise in quantum circuits, providing a more accurate noise control method for quantum computing.

[0060] 3. Improved classification accuracy and system performance:

[0061] After training and optimizing the quantum anti-noise model, the classification accuracy has been significantly improved, proving the effectiveness of the anti-noise model on the actual data set. This was effectively verified by comparing the classification accuracy and loss curves before and after adding the anti-noise model.

[0062] 4. Combining traditional deep learning algorithms with quantum computing advantages:

[0063] By combining the quantum anti-noise model with the traditional deep learning algorithm, the present invention not only utilizes the advantages of deep learning in processing large-scale data, but also gives full play to the high efficiency of quantum computing in processing complex computing tasks, thereby realizing a new and efficient method for resisting quantum noise.

[0064] This invention effectively solves the noise problem in quantum computing through innovative technical methods, improves the practicality and accuracy of quantum machine learning models, and is an important technological advancement in the field of quantum computing.

[0065] Third, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

[0066] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0067] ① Lowering the technical threshold: By simulating QCNN models containing quantum noise, research institutions and companies that do not have advanced quantum hardware can also participate in and benefit from quantum computing.

[0068] ② Commercialization potential: With the development of quantum computing, the demand for anti-noise technology will increase. When the technology matures, it can form new market demand and promote the development and commercialization of related anti-noise products and services.

[0069] ③Expected benefits: By reducing the impact of noise, quantum technology can be applied to scenarios where reliable results cannot be obtained due to excessive noise, such as complex chemical reaction simulations or financial market analysis.

[0070] (2) The technical solution of the present invention fills the technical gap in the industry at home and abroad:

[0071] The technical solution of the present invention successfully fills the technical gap in the field of quantum computing at home and abroad through its original quantum random demotion technology. Previously, the methods of managing and reducing noise in quantum systems were either not effective enough or too complicated and costly. The present invention provides a practical and economical solution to this problem, which is an area that has not been explored or realized in the industry. With this technology, we can expect to see the widespread deployment of quantum computing in practical applications, especially those applications that are particularly sensitive to noise. Therefore, this invention not only represents a major technological advancement, but also lays the foundation for the application and commercialization of quantum computing, which will have a far-reaching impact on the entire field of quantum computing. "

[0072] (3) The technical solution of the present invention solves the technical problems that people have been eager to solve but have never been able to solve successfully:

[0073] The technical solution of the present invention, i.e. the successful application of the method of resisting quantum noise interference based on quantum random dethronement in large-scale quantum circuits, quantum convolutional neural networks (QCNN), solves the technical problem that people have been eager to solve but have not been effectively overcome so far, i.e., the problem of reducing the impact of quantum noise on calculation results in quantum circuits. This problem has always been one of the most challenging problems in the field of quantum computing and an important factor limiting the commercialization and widespread application of quantum computing.

[0074] (4) The technical solution of the present invention overcomes technical prejudice:

[0075] The technical solution of the present invention has been carefully designed to overcome the bias existing in existing technologies to a certain extent, providing more fair and balanced calculation results. Through comprehensive testing and data analysis, we ensure that the solution can maintain consistent performance in different situations and is not affected by preconditions and external assumptions. In addition, this technical solution also focuses on reducing the user's usage threshold to ensure that users with different backgrounds and abilities can benefit from it, avoiding the knowledge or resource bias that is prone to occur in previous technologies. Therefore, the technical solution of the present invention is not only innovative at the technical level, but also contributes to ensuring technical fairness.

[0076] Fourthly, the quantum noise interference resistance system based on quantum random dethronement provided by the present invention proposes an effective solution to the noise interference problem in current quantum computing, and has the following significant technical advances:

[0077] (1) Enhanced anti-noise capability: Through quantum random demotion technology, the system can effectively resist interference from quantum noise during quantum computing, thereby improving the robustness and reliability of the system.

[0078] (2) Data preprocessing and coding optimization: The data preprocessing and coding modules in the system can convert classical data into quantum data and use quantum angle encoding to better adapt to the quantum computing platform and improve the accuracy and availability of data.

[0079] (3) Quantum entanglement and measurement module: Through quantum entanglement technology, the system can realize the entanglement of quantum data in the quantum convolutional neural network. At the same time, the quantum measurement module can convert quantum data into classical data, providing a reliable foundation for subsequent data processing.

[0080] (4) Anti-noise module optimization: The anti-quantum noise module in the system can effectively combat the adverse effects of quantum noise on the classifier. By evaluating the classification accuracy and loss function of the model, the hyperparameters of the model are optimized, thereby improving the performance and robustness of the system.

[0081] In summary, the quantum noise interference resistance system based on quantum random demotion provided by the present invention has important technical significance in the field of quantum computing, effectively solves the noise interference problem in quantum computing, and provides reliable support and guarantee for the practical application of quantum computing. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a flow chart of a method for resisting quantum noise interference based on quantum random dethronement provided by an embodiment of the present invention.

[0083] Figure 2 It is a schematic diagram of a method for resisting quantum noise interference based on quantum random dethronement provided by an embodiment of the present invention.

[0084] Figure 3 This is a diagram of a quantum convolution layer model provided by an embodiment of the present invention.

[0085] Figure 4 It is a diagram of a quantum anti-noise model provided by an embodiment of the present invention.

[0086] Figure 5 3 is an out-of-sample error graph under different Layer elimination rates and Gate elimination rates provided by an embodiment of the present invention.

[0087] Figure 6 It is a structural block diagram of a system for resisting quantum noise interference based on quantum random demotion provided by an embodiment of the present invention.

[0088] Figure 7 This is the effect of applying the method for resisting quantum noise interference based on quantum random dethronement provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0090] like Figure 1 As shown, a method for resisting quantum noise interference based on quantum random dethronement provided by an embodiment of the present invention includes the following steps:

[0091] S101, data collection, downloading MNIST dataset and Fashion-MNIST dataset;

[0092] S102, extracting required training samples from the MNIST dataset and the Fashion-MNIST dataset and shuffling the order of the samples;

[0093] S103, preprocessing the data, including image downsampling and normalization;

[0094] S104, build a quantum noise model and deploy it on a quantum convolutional neural network;

[0095] S105: Establish a quantum convolutional neural network model containing quantum noise and train the data set;

[0096] S106: Build a quantum noise-resistant model and deploy it on a noisy quantum convolutional neural network;

[0097] S107: Train the quantum noise-resistant model by adjusting different noise intensities;

[0098] S108: By comparing the classification accuracy and loss curve evaluation model before and after the addition of the quantum anti-noise model, the quantum anti-noise model and the traditional deep learning algorithm are combined to achieve resistance to noise in quantum circuits.

[0099] The following is an explanation of the detailed working principle of a method for resisting quantum noise interference based on quantum random dethronement provided by an embodiment of the present invention:

[0100] S101, data collection:

[0101] In this step, download the MNIST dataset and Fashion-MNIST dataset. MNIST is a dataset of handwritten digits, including images of handwritten digits from 0 to 9 and their corresponding labels; while Fashion-MNIST is a dataset of fashion products, including images of fashion products in 10 categories and their corresponding labels. These datasets provide rich training samples for subsequent model training.

[0102] S102, extract training samples and shuffle them:

[0103] Extract the required training samples from the downloaded data set and shuffle the order of these samples. The purpose of this step is to avoid overfitting of the model during training and improve the generalization ability of the model.

[0104] S103, data preprocessing:

[0105] In this step, the extracted training samples are preprocessed. Preprocessing includes image downsampling and normalization. Image downsampling can reduce the size of the image and reduce the amount of model calculation; normalization is to adjust the data to a certain range, such as [0,1] or [-1,1], to facilitate model training.

[0106] S104, build a quantum noise model and deploy it on a quantum convolutional neural network:

[0107] In this step, a quantum noise model is constructed based on the actual quantum hardware environment or theoretical model. This model will simulate the noise interference in the quantum circuit. Then, this noise model is deployed on the quantum convolutional neural network (QCNN) to simulate the impact of quantum noise on model performance in the subsequent training process.

[0108] S105, establish a quantum convolutional neural network model containing quantum noise and train it:

[0109] After deploying the quantum noise model on QCNN, a QCNN model with quantum noise is obtained. Next, the model is trained using the preprocessed MNIST and Fashion-MNIST datasets. The purpose of the training is to enable the model to accurately identify the categories in the image under noise interference.

[0110] S106, build a quantum noise-resistant model and deploy it on a noisy quantum convolutional neural network:

[0111] While training the noisy QCNN model, build a quantum noise-resistant model. The goal of this model is to reduce or eliminate the impact of quantum noise on the performance of QCNN. Once built, deploy the quantum noise-resistant model on the noisy QCNN.

[0112] S107, train the quantum noise-resistant model by adjusting different noise intensities:

[0113] In this step, the quantum noise-resistant model is trained by adjusting the noise intensity of the quantum noise model. In this way, the quantum noise-resistant model can adapt to noise interference of different intensities and improve its robustness in practical applications.

[0114] S108, Evaluate the model and achieve resistance to quantum noise:

[0115] Finally, the performance of the model is evaluated by comparing the classification accuracy and loss curve before and after adding the quantum anti-noise model. If the classification accuracy of the model is improved after adding the anti-noise model, and the loss curve also shows a better convergence trend, then it means that the method can effectively resist the interference of quantum noise. Finally, the quantum anti-noise model is combined with the traditional deep learning algorithm to achieve noise resistance in quantum circuits in practical applications.

[0116] The method provided in the embodiment of the present invention improves the performance of quantum convolutional neural network under noise interference by constructing a quantum noise model and a quantum anti-noise model, thereby providing a guarantee for the stability and reliability of quantum computing in practical applications.

[0117] like Figure 6 As shown, the present invention also provides a system for resisting quantum noise interference based on quantum random dethronement, including:

[0118] ① Data collection module, downloading MNIST dataset and Fashion-MNIST dataset from the Internet and importing them into the experimental platform;

[0119] ②Data preprocessing module: used to cut and reduce the dimension of data set images to meet the data input requirements of the quantum platform;

[0120] ③Data encoding module: used to encode classical data into quantum data, using quantum angle encoding, and the encoded data participates in subsequent calculations and training;

[0121] ④Quantum entanglement module: used to entangle quantum data, realize quantum data entanglement in the constructed quantum convolutional neural network, and train image classification tasks;

[0122] ⑤Quantum noise module: used to establish a quantum noise model and deploy it on a quantum convolutional neural network to realize the experimental environment of a noisy quantum neural network;

[0123] ⑥Quantum measurement module: used to measure quantum data and convert it into classical data, so as to facilitate training and updating of the subsequent data post-processing module;

[0124] ⑦Quantum post-processing module: used to train and optimize the classical data after measurement, using the gradient descent method to train to achieve the desired effect;

[0125] ⑧Classification module: used to classify classic data and achieve the expected image classification task.

[0126] ⑨ Anti-quantum noise module: used to combat the adverse effects of quantum noise on the classifier. The model is evaluated by two indicators: classification accuracy and loss function. The out-of-sample error is used to analyze the effects of different hyperparameters on the anti-noise model.

[0127] A computer device provided by an embodiment of the present invention includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for resisting quantum noise interference based on quantum random demotion.

[0128] A computer-readable storage medium provided by an embodiment of the present invention stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for resisting quantum noise interference based on quantum random demotion.

[0129] An embodiment of the present invention provides an information data processing terminal, which is used to implement the quantum random dethronement-based anti-quantum noise interference system.

[0130] The detailed working principle of the present invention is as follows:

[0131] First, the data acquisition module downloads the MNIST dataset and Fashion-MNIST dataset from the Internet and imports them into the experimental platform. Both datasets are image classification datasets used to train and test image classifiers.

[0132] Next, the data preprocessing module performs segmentation and dimensionality reduction operations on the imported data set images. This is because the quantum computing platform has specific requirements for the dimension and format of the input data, so it is necessary to convert the data into a format acceptable to the quantum computing platform through preprocessing.

[0133] Then, the data encoding module encodes the classical data into quantum data. Here, quantum angle encoding is used to map the classical data to quantum states so that it can be processed on the quantum computing platform. The encoded data will participate in the subsequent calculation and training process.

[0134] The quantum entanglement module is used to entangle quantum data. In quantum convolutional neural networks, by entangling quantum data, the correlation between data can be enhanced and the accuracy of classification can be improved. The entanglement operation is implemented on a quantum computing platform and can be used to train image classification tasks.

[0135] In order to simulate quantum noise in a real environment, the quantum noise module establishes a quantum noise model and deploys it on a quantum convolutional neural network. In this way, the entire network can be trained and tested in a noisy environment to evaluate the impact of noise on classifier performance.

[0136] During the quantum computing process, the quantum measurement module is responsible for measuring quantum data and converting it into classical data. This is because the results of quantum computing need to be output in the form of classical data for subsequent processing and analysis.

[0137] The quantum post-processing module trains and optimizes the measured classical data. The gradient descent method is used for training, and the performance of the classifier is optimized by continuously adjusting the network parameters. The goal of the optimization is to improve the classification accuracy and reduce the loss function value.

[0138] The classification module classifies the optimized classical data to achieve the expected image classification task. Based on the classification results, the performance of the entire system can be evaluated.

[0139] Finally, the anti-quantum noise module is used to combat the adverse effects of quantum noise on the classifier. The quality of the model can be evaluated by evaluating the classification accuracy and loss function. At the same time, the module also analyzes the effects of different hyperparameters on the anti-noise model through out-of-sample errors, so as to further optimize the performance of the model.

[0140] As for the computer device and the computer readable storage medium, they store computer programs for implementing the method of the present invention. When these programs are executed by the processor, they will operate according to the above working principle to implement the method of resisting quantum noise interference based on quantum random dethronement.

[0141] The information data processing terminal is used to realize the entire quantum noise interference resistance system based on quantum random dethronement. It integrates functional modules such as data acquisition, preprocessing, encoding, entanglement, noise simulation, measurement, post-processing, classification and anti-noise, providing users with a complete image classification solution.

[0142] In summary, the present invention uses quantum computing and quantum entanglement technology, combined with an anti-quantum noise module, to achieve the optimization of noisy quantum neural networks and the execution of image classification tasks. The entire system is efficient, accurate, and noise-resistant, providing strong support for the application of quantum computing in image classification and other fields.

[0143] The technical solution of the present invention is further described below in conjunction with embodiments.

[0144] Example 1

[0145] like Figure 2 As shown, the technical solution of the present invention includes:

[0146] Step 1: Data collection: download the MNIST dataset and Fashion-MNIST dataset;

[0147] Step 2: Extract the required training samples from the MNIST dataset and Fashion-MNIST dataset and shuffle the sample order;

[0148] Step 3: preprocess the data, including image downsampling and normalization;

[0149] Step 4: Build a quantum noise model and deploy it on a quantum convolutional neural network. The quantum noise model uses depolarization channels, quantum bit flips, quantum phase flips and other methods to convert pure quantum bits into mixed quantum bits, and introduces the concept of quantum noise intensity. The quantum noise intensity is achieved by adjusting the frequency of different depolarization channels, quantum bit flips, and quantum phase flips in the circuit to control the strength of the noise in the quantum circuit. The greater the quantum noise intensity, the more quantum bits in the circuit will have errors due to noise interference, and the more obvious the classification effect will be.

[0150] Step 5: Establish a quantum convolutional neural network model containing quantum noise and train the data set; the model includes a quantum convolution layer, a pooling layer, a fully connected layer, a quantum noise module, and a data post-processing module; the quantum convolution layer is constructed using quantum circuits, and the required quantum circuits are shown in Table 1. The quantum convolution layer realizes the entanglement of quantum data. Previous studies have shown that the entanglement characteristics of quantum are an important reason for improving the expressibility of circuits to achieve high classification accuracy. Therefore, a controlled NOT gate (CNOT) or a Pauli-Z gate with entanglement characteristics is combined to realize the quantum convolution layer. The specific quantum convolution layer is as follows Figure 3 shown.

[0151] Step 6: Construct a quantum noise-resistant model and deploy it on a noisy quantum convolutional neural network; the quantum noise-resistant model mainly uses the method of quantum random elimination. Random elimination refers to the removal of random quantum entangled gate circuits (such as CNOT gates, Pauli-Z gates with entanglement, etc.) with a certain probability. By adjusting the layer elimination rate and the gate elimination rate, the elimination rate of each layer of the quantum convolution layer and the elimination rate of each entangled gate in each layer are controlled to reduce the expressibility of the quantum circuit, so as to deal with the problem of reduced classification accuracy caused by the quantum circuit being too "sensitive" under the influence of noise. The specific implementation method is as follows Figure 4 As shown;

[0152] Step 7: Train the quantum noise-resistant model by adjusting different noise intensities;

[0153] Step 8: After quantum evaluation, the classification accuracy and loss curve before and after the addition of the quantum anti-noise model are compared, and the quantum anti-noise model is combined with the traditional deep learning algorithm to resist the noise in the quantum circuit.

[0154]

[0155] Table 1

[0156] Example 2

[0157] For the entire noisy quantum convolutional neural network, the quantum convolution circuit and quantum gate circuit are built with the Pennylane framework; the pooling layer and the fully connected layer are built with the Pytorch framework, and the classifier is the Adagrad classifier. At the beginning of the iteration, the learning rate is set to a large value. After every 10 epochs, the learning rate is reduced by half, so that the model can stably find the optimal point.

[0158] The present invention is dedicated to solving the problem that the quantum-classical hybrid neural network in the past was insufficient in considering quantum noise, and that the image recognition and classification accuracy of large-scale quantum circuits (such as quantum convolutional neural networks) was seriously reduced under the interference of quantum noise. The invention proposes a new experimental method to create an experimental environment by simulating noisy quantum circuits using classical computers. This method makes the results of the quantum neural network model running in a classical computer environment basically consistent with the experimental results conducted in a real quantum device environment, so the experimental results are more practical. In addition, the invention solves the problem of resisting noise interference in large-scale quantum circuits, which makes image recognition possible in the noise-in-scale quantum (NISQ) era.

[0159] For each data set, the sample size is 1200, of which 1000 are used for training samples and 200 are used for testing to verify the accuracy of the algorithm. To enhance the stability of the experimental results, each experiment was repeated 10 times, and the average of the results of these 10 experiments was used as the final loss value and accuracy. The loss value was calculated using the cross entropy loss function. Figure 5 It can be seen that under qualitative analysis, a lower layer eliminate rate combined with a higher gate eliminate rate can achieve better anti-noise effect.

[0160] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0161] In order to prove the creativity and technical value of the technical solution of the present invention, this section provides application examples of the claimed technical solution on specific products or related technologies.

[0162] Embodiment 1:

[0163] Image recognition software products that use this patented technical solution. The software has a built-in quantum-classical hybrid quantum convolutional neural network (QCNN), and adds a method to resist quantum noise interference based on quantum random dethronement during the training process. When implementing the image classification and recognition functions, first, the image data is feature extracted and classified through the quantum convolutional neural network, and then the error is corrected through the built-in noise attenuation processing part to further improve the accuracy of image recognition. In actual tests, this image recognition software has significantly improved its performance in resisting quantum noise compared to traditional image recognition software.

[0164] Embodiment 2:

[0165] An intelligent security system using the patented technical solution. The system integrates an intelligent image analysis module consisting of a quantum convolutional neural network and a quantum noise interference resistance system based on quantum random dethronement. This module can identify various objects (such as pedestrians, vehicles, etc.) in the monitoring screen through a quantum convolutional neural network, and then eliminate possible errors through a quantum noise interference resistance system based on quantum random dethronement, and finally achieve high-accuracy monitoring within the monitoring range. Compared with traditional security systems, this intelligent security system can achieve a higher recognition rate in the presence of quantum noise interference.

[0166] The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the prior art. The following content is described in conjunction with data, charts, etc. of the test process.

[0167] In an ideal (no quantum noise) environment, a three-layer quantum convolutional neural network can achieve an accuracy of more than 95% for the three-class classification of the MNIST dataset. Once quantum noise is added, the accuracy will drop by 10% to 85%. Currently, there is no effective method for quantum convolutional neural networks to deal with the interference caused by this noise. The method of resisting quantum noise interference based on quantum random dethronement proposed in this invention can improve the accuracy to 92% after experimental testing, which is very close to the effect under an ideal environment. The experimental results are as follows Figure 7 shown.

[0168] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for resisting quantum noise interference based on quantum random dethronement, characterized in that: The following steps are involved: Step 1: Data collection, downloading the data set; Step 2: extract the required training samples from the data set and shuffle the sample order; Step 3: preprocess the data, including image downsampling and normalization; Step 4: Build a quantum noise model and deploy it on a quantum convolutional neural network. Step 5: Establish a quantum convolutional neural network model containing quantum noise and train the data set; Step 6: Build a quantum noise-resistant VQC and deploy it on a noisy quantum convolutional neural network. Step 7: Train the quantum noise-resistant model by adjusting different noise intensities; Step 8: By comparing the classification accuracy and loss curve evaluation model before and after the addition of the quantum noise-resistant model, the quantum noise-resistant VQC and traditional deep learning algorithms are combined to resist the noise in the quantum circuit. In step 1, download the MNIST dataset and Fashion-MNIST dataset. MNIST is a dataset of handwritten digits, including images of handwritten digits from 0 to 9 and their corresponding labels. Fashion-MNIST is a dataset of fashion products, including images of fashion products in 10 categories and their corresponding labels. These datasets provide abundant training samples for subsequent model training. The quantum noise-resistant VQC in step 6 is inspired by the idea of ​​deep learning. It uses a certain strategy to randomly remove quantum gates in the circuit to reduce the expressibility of the quantum convolutional neural network. The strategy includes: (1) Randomly remove the CNOT gate or Pauli-Z gate containing entanglement in the quantum circuit during the i-th parameter update; (2) Calculate the gradient of the loss function generated after removing the entangled quantum gate, determine the i+1th parameter by the gradient descent method, and update the parameter; (3) Repeat the above steps until the model accuracy reaches the expected value.

2. The method for resisting quantum noise interference based on quantum random dethronement as claimed in claim 1, characterized in that: In step 4, the quantum noise model adds noise to the ideal quantum circuit to make the experimental environment more consistent with the experimental environment of an actual quantum computer, including: (1) Use bit flip, phase flip, and depolarizing channel methods to simulate the situation when the quantum bit is disturbed by quantum noise; (2) The quantum data affected by quantum noise is transformed into classical data through the subsequent quantum measurement module.

3. The method for resisting quantum noise interference based on quantum random dethronement as claimed in claim 1, characterized in that: The quantum convolutional neural network model containing quantum noise in step five includes the following seven groups of modules: data preprocessing module, quantum coding module, quantum entanglement module, quantum noise module, quantum measurement module, classification module, and data post-processing module.

4. The method for resisting quantum noise interference based on quantum random dethronement as claimed in claim 3, characterized in that: The function of the data preprocessing module is to use dimensionality reduction, downsampling, and data flattening methods to process input data to meet quantum requirements; the function of the quantum coding module is to use an angle coding method, and each channel is encoded with a quantum bit to realize the conversion of classical data to quantum state; the quantum entanglement module is to form a quantum convolution layer through Hadamard gates, single-qubit rotation gates, and CNOT gates, and perform unitary transformation on the encoded quantum data to realize the transformation and entanglement of quantum states; the function of the quantum measurement module is to measure the transformed and entangled quantum state to obtain the expected value with Hamiltonian, and pass the result to the classical post-processing module.

5. The method for resisting quantum noise interference based on quantum random dethronement as claimed in claim 3, characterized in that: The function of the data post-processing module is to perform gradient training on the obtained expectation and the defined loss function, and dynamically update the entanglement module training parameters.

6. A system for resisting quantum noise interference based on quantum random dethronement, which implements the method for resisting quantum noise interference based on quantum random dethronement as claimed in any one of claims 1 to 5, characterized in that: The quantum noise interference resistance system based on quantum random dethronement includes: (1) Data collection module, used to download the MNIST dataset and Fashion-MNIST dataset from the Internet and import them into the experimental platform; (2) Data preprocessing module, used to cut and reduce the dimension of the data set images to meet the data input requirements of the quantum platform; (3) Data encoding module, used to encode classical data into quantum data, using quantum angle encoding, and the encoded data participates in subsequent calculations and training; (4) Quantum entanglement module, used to entangle quantum data, realize quantum data entanglement in the constructed quantum convolutional neural network, and train the image classification task; (5) Quantum noise module, which is used to establish a quantum noise model and deploy it on a quantum convolutional neural network to realize an experimental environment for noisy quantum neural networks; (6) Quantum measurement module, used to measure quantum data and convert it into classical data, so as to facilitate training and updating of the subsequent data post-processing module; (7) Quantum post-processing module, which is used to train and optimize the classical data after measurement, using the gradient descent method to achieve the desired effect; (8) Classification module, used to classify classical data and achieve the expected image classification task; (9) Anti-quantum noise module is used to combat the adverse effects of quantum noise on the classifier. The model is evaluated by two indicators: classification accuracy and loss function. The out-of-sample error is used to analyze the effects of different hyperparameters on the anti-noise model.

7. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for resisting quantum noise interference based on quantum random demotion as described in any one of claims 1 to 4.

8. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for resisting quantum noise interference based on quantum random demotion as described in any one of claims 1 to 4.

9. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the quantum noise interference resistance system based on quantum random dethronement as described in claim 6.

Images