Research On Data Classification Algorithm Based On Quantum Mechanics Theory

Classification algorithms are machine learning methods that classify samples from unknown categories by training and learning based on a training sample set of known categories.The algorithms are theoretically robust and can efficiently obtain valuable information from the training sample set,and have strong data processing capabilities.which has been broadly used in text classification,image recognition and other fields.Whereas,account for the enormous volume of data in today’s society,classical classification algorithms suffer from excessive storage space requirements and high time complexity.The emergence of quantum computing has spawned a new research direction of classification-quantum classification.The novel algorithms can effectively solve the dilemma of the current classical classification algorithms.The quantum classification algorithms is based on the framework of the classical classification algorithms,using the superposition and parallelism of quantum computing to design quantum algorithms and circuits,and realize the quantum versions of classical classification algorithms to achieve the effect of acceleration.Over the years,scholars have studied quantum classification algorithms and proposed a series of effective quantum classification algorithms.On this basis,this paper investigates the quantum Bayesian classification algorithm and quantum-nearest neighbor classification algorithm,and proposes two quantum classification algorithms with accelerated effect,the details of this paper are as follows.1.A Bayesian binary classification algorithm based on quantum counting is proposed.Firstly,the required quantum states are prepared through quantum random access memory,and the oracle black box operation is used to phase flip and construct the corresponding operator;secondly,the quantum states are redescribed on the eigenstate space of the operator,and the phase estimation is performed with the help of auxiliary particles;finally,the projection measurement is used to efficiently obtain the data required for Bayesian classification,and the quantum Bayesian binary classification is achieved.The theoretical analysis shows that the algorithm has exponential speed up in low-dimensional feature space compared with the classical algorithm.In addition,this paper applies the above algorithm to the news text classification problem and gives the corresponding classification procedure.2.A quantum -nearest neighbor classification algorithm is designed,which is based on the Mahalanobis distance.The quantum-nearest neighbor classification algorithm distance metric is implemented by quantum phase estimation and amplitude estimation to encode the distance information onto the quantum states.Here,the quantum Hamiltonian simulation method is used to model the covariance matrix,which has exponential acceleration in time complexity compared to the classical method.Based on this,the quantum technique is used to construct -minimum search algorithm for disordered data sets,which searches the training samples closest to the test sample and obtains the class with the highest frequency among them as the class of the test sample.The algorithm improves the classification efficiency by parallelising the distance calculation and reducing the complexity of the search process.Compared with the classical algorithm,the quantum -nearest neighbor classification algorithm based on the Mahalanobis distance proposed in this paper has the effect of quadratic acceleration.