Research on Clustering Based on graph neural network

Abstract:
In the past few years, the rise and application of neural network, it has successfully promoted the research of pattern recognition and data mining, become the focus of attention. Many machine learning tasks that used to rely heavily on manual feature extraction, today, it has been completely changed by the method of deep learning. Although the traditional deep learning method has achieved great success in extracting the features of Euclidean space data, the data in many practical application scenarios are generated from non Euclidean space, the performance of traditional deep learning methods in dealing with non European spatial data is still unsatisfactory, for Non-European spatial data, it is necessary to analyze the hidden structural information combined with the relevant knowledge of graph theory. Therefore, the concept of graph neural network is introduced to improve the traditional clustering algorithm, so that the original data structure information can be retained as much as possible while processing high-dimensional complex data, so as to achieve the goal of optimizing clustering.