Data processing method and device, computer equipment and storage medium

Abstract

The invention discloses a data processing method and device, computer equipment and a storage medium, and belongs to the technical field of computers. In the parameter adjustment process of the graph neural network, the conflict level parameter of each labeled node is determined to measure the topological position of each labeled node, on the basis of the conflict level parameter, the target weight of each labeled node is allocated to the labeled node, and the target weight is put into the parameter adjustment process, so that the topological position of each labeled node is measured. Influences generated by different labeling nodes at different topological positions in the parameter adjustment process are adjusted, for example, large target weights are distributed to labeling nodes with topological positions close to the center of a category, and small weights are classified to labeling nodes with topological positions close to the boundary of the category. Therefore, the phenomenon of category imbalance generally existing in the graph neural network can be improved, and the recognition accuracy of the graph neural network is improved.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to a data processing method, device, computer equipment and storage medium. Background technique [0002] With the development of computer technology, although traditional deep learning methods have been applied to extract features of Euclidean spatial data with great success, the data in many practical application scenarios (such as graph data) are generated from non-Euclidean spaces. , therefore, Graph Neural Networks (Graph Neural Networks, GNN) has gradually become a research hotspot. A graph neural network refers to a neural network structure used to process graph data, where graph data refers to a data structure including nodes and edges. For example, each account in a social network may correspond to a node in a graph neural network. When When account A and account B have a friendship relationship, there is a connected edge between the corresponding nodes of acco...

AI Technical Summary

Problems solved by technology

Because in the training phase of the graph neural network, the number of label nodes contained in the label sets provided for different categories is not the same, and the position distribution of the label nodes contained in each label set in the graph neural network is also different. Balanced, leading to obvious differences in the prediction ability of graph neural network for different categories when processing node classification tasks, for example, the recognition accuracy for category 1 is generally high, but the recognition accuracy for category 2 is generally low

Images