Research On Key Technologies Of Quantum Algorithm For Data Mining

As an emerging technology of computation,quantum computing has received much attention for its incomparable computing capabilities.Especially,by the introduction of some novel quantum algorithms,quantum computing has performed superiorly in the field of information processing.To further accelerate the implementation of quantum technologies,lots of medium to long term research plans in this field have been promulgated by many countries in the world.Data mining is widely used technology to extract the implicit information among the massive data.And a series of practical applications have been constructed on it,for example search engines and social network,which greatly impacted our lives.However,with the rapid increase of data volume,the limited processing abilities has become the biggest bottleneck to deal with the amounts of data.And quantum computing seems be a new breakthrough for this dilemma benefited from its outstanding performance,attracting the researchers to explore how to apply quantum technology to traditional data mining.In this thesis,we will study some problems in the field of data mining based on the quantum computation.Specifically,the main contents of this article are as following:1.Summarize the theories and methods of current quantum computing used in the field of data mining.During the plenty of theories in quantum technology,in this paper we will focus on the basic theory of quantum computing applied to data mining field,and explain its excellent property varied from classic computing.In addition,the general rules of algorithm design are also deducted by us to provide important references for subsequent related research.2.Research on image matching based on quantum computingImage matching is the critical in digital image processing,which plays an important role in target positioning and medical image analysis.In this paper,a new quantum image representation model is designed based on the entanglement and superposition characteristics of qubits,and then the algorithm of quantum image matching based on the model is implemented.Compared with the classic image matching algorithm,the algorithm can achieve a square-level acceleration in performance.3.Research on graph isomorphism based on quantum computingAs one of the well-known problems in the classical graph processing field,even though many arduous efforts have been dedicated in this problem,graph isomorphism still can not be resolved during polynomial time complexity.But quantum computing has contributed the new solutions for this.In this article,we use quantum theory to design a practical approximate solution algorithm on graph isomorphism.Experiments have found that this method performed well on some graph datasets than other graph isomorphism solutions.4.Graph similarity measurement method based on quantum computingGraph similarity measurement actors essentially in classification and clustering of the graph data structures,but traditional methods perform poorly in accuracy and complexity.To improve,we design a new graph kernel function based on quantum technology,which uses the quantum walk model to extract topological features.By experiment we can verify that this method achieved better in the graph classification task compared with some existing classical graph kernel methods.5.Node centrality measurement method based on quantum computingFor the detection of node centrality,most existing solutions directly measure the centrality based on the graph topology.In this paper,we try to take the reversible characteristic of quantum computing and then use the inverse process of quantum walk to explore the centrality of node in the graph.At the same time,we found that this measurement method can also be exploited to the community discovery task to optimize the cold start and community annexation problems in the classic label propagation algorithm,and experiments indicated that the optimized algorithm performed more accurately and steadily among the community division tasks.In this paper,we aimed at some difficult problems in the data mining field and try to explore the executable solutions by the utilization of quantum characteristics.And the series of application issues are also explored to demonstrate the outstanding performance and wide application of quantum technologies.Furthermore,the researches in this paper can not only promote the cross-fusion of quantum computing and data mining,but also provide effective guidance for the popularization and application of quantum computing in other information processing fields.