Attribute graph deep clustering method of hierarchical graph convolutional network based on attention mechanism

Abstract

The invention relates to an attribute graph deep clustering method based on a graph convolutional network, which considers the necessity of adaptively and hierarchically learning neighbor nodes of the graph convolutional network, proposes a hierarchical graph convolutional network based on an attention mechanism, and solves the defect that the graph convolutional network is insufficient in learning surrounding neighbor node information. According to the method, the hidden layers are iterated for the graph convolutional network, so that the graph convolutional network can learn multi-hop neighbor information, an attention mechanism is added into the first hidden layer, so that the graph convolutional network can adaptively learn first-order neighbor information, and the first-order neighbor information is closer and more important to nodes. Through analysis of an experimental result, the improvement can significantly improve the clustering precision for a deep clustering task of the attribute graph.

Description

technical field [0001] The technical solution of the present invention relates to the fields of graph convolutional network and deep clustering, specifically, an attribute graph clustering method of hierarchical graph convolutional network based on attention mechanism. Background technique [0002] Many applications in the real world, such as Web networks and social networks, can be described by graph structures, or can be processed using methods based on graph data. The graph-based representation characterizes individual attributes through node attributes, while capturing pairwise relationships through the graph structure. However, the scale of graph data is increasingly trending towards massive development. On the one hand, analyzing the information users need from these large-scale graphs is a very challenging problem. On the other hand, such a large-scale Figure 1 Loading and displaying at one time not only consumes a lot of resources, but also makes the visual view v...

AI Technical Summary

Problems solved by technology

[0006] (2) The traditional deep clustering based on graph convolutional network does not consider the weight influence of graph convolutional network on the learning of surrounding neighbors

Images