Graph classification method based on quantum walk

Abstract

The invention provides a graph classification method based on a quantum walk technology aiming at the problems of high difficulty in graph data analysis and prediction, low classification precision and the like. Feature extraction calculation is performed on a specific substructure in graph data by utilizing quantum walk amplitude, so that efficient isomorphic substructure matching is realized, and graph similarity measurement is completed. The method specifically comprises a graph kernel construction method based on a quantum walk method, and the method comprises the following steps: firstly, carrying out the multi-step discrete time quantum walk of each graph, completing the construction of a graph kernel through the matching of a substructure through a neighborhood pair based on the rapid quantum walk amplitude calculation, and finally, in combination with a kernel-based support vector machine, training to obtain a graph classifier. The graph structure data in the real world can be efficiently analyzed, and the category of unknown new data can be accurately judged.

Description

technical field [0001] The invention relates to a graph classification method based on quantum walk, which mainly uses a new quantum computing method to realize high-precision intelligent learning, analysis and prediction for graph data. Background technique [0002] Graph is an important data structure in the field of information processing. It can clearly express the connection relationship between individuals in the real world. It is widely used in many fields such as social network, biomedicine, machine vision and program analysis. The problem of graph classification refers to the study of class judgments on graphs through structural analysis. This problem is of great value in practical applications. For example, in the field of chemical research, using the comparison of topological features of molecular structures, researchers can make predictions about the chemical properties of unknown molecules. For another example, in the field of program design, through comparati...

AI Technical Summary

Problems solved by technology

However, the existing graph kernel-based classification methods are all approximations to the accurate classification methods based on subgraph isomorphism, and they all make a corresponding compromise in terms of computational complexity and classification accuracy.

Although many mature graph kernels have appeared, there are still some problems: 1) Since the relative position information between substructures cannot be constructed, the classification accuracy is affected when classifying based on these graph kernels

2) Since graph attributes are difficult to obtain in practice, these graph checks are not capable of expressing graph similarity enough

3) These graph kernels cannot distinguish some similar but non-isomorphic graphs, such as cospectral graphs and regular graphs, etc.

[0005] At present, the problem of low classification accuracy in the graph classification problem exists, and the quantum walk technology can effectively solve these problems

Images