Multi-view comparative learning-based citation network graph representation learning system and method

Abstract

The invention discloses a multi-view comparative learning-based citation network graph representation learning system and method. The citation network graph representation learning system comprises: a sample construction module which takes original graph node representation as a positive sample and constructs a negative corresponding sample based on an original graph; a graph enhancement module which is used for enhancing the positive sample node features based on a personalized page ranking algorithm and a Laplace smoothing algorithm to obtain a positive sample graph and a negative sample graph; a fusion module which is used for extracting the positive sample graph representation and the negative sample graph representation based on an encoder, integrating the positive sample graph representation and the negative sample graph representation, and obtaining the consensus representation of the positive sample graph and the negative sample graph through a cross view concentrated fusion layer; a mutual information estimation module which is used for comparing the learning representations of the positive sample pair and the negative sample pair through a discriminator; and a difficult sample mining module which represents the consistency between the negative sample pairs according to a pre-calculated affinity vector, and selects and retains nodes which are difficult to express global or neighbor information.

Description

technical field [0001] The invention relates to the technical field of citation network-oriented graph representation learning, in particular to a citation network graph representation learning system and method based on multi-view comparison learning. Background technique [0002] With the development of informatization, the continuous increase of storage space, and the continuous expansion of Internet user groups, a large amount of literature information is generated every day. How to use these literatures to better serve different technical industries and a large number of groups has become the focus of researchers. Content. A citation network refers to a collection of citation and cited relationships between documents, including scientific journals, patent documents, conference proceedings, scientific reports, and dissertations. A large amount of English data exists in various applications in real life. By analyzing it in different dimensions, it can help users understa...

AI Technical Summary

Problems solved by technology

This mutual information evaluation method has the following two problems. 1) It only pays attention to the graph structure information from a single perspective, ignoring the position information of subgraph structures in different ranges; 2) it extracts all node information indiscriminately, making difficult samples Difficult to distinguish

Through the above method, the network usually prefers to fit the overall or extreme local graph representation, and it will cause confusion when identifying difficult nodes, which will lead to inaccurate estimation of the similarity between node representations, and the learned graph representation recognition ability is limited.

Images