A Topic-Based Network Representation Learning Method

Abstract

The invention discloses a topic-based network representation learning method, which belongs to the technical field of network representation learning. The method of the present invention includes: designing a self-encoder model, defining the model structure and the form of input and output data, so as to retain the structural features represented by the network topology; utilizing node information in the network to expand the self-encoder network model , extract the information contained in the nodes in the network, and integrate the topic factors into it to retain the semantic features represented by it, while retaining the structural features of the global network; the two types of features are fused into the low-dimensional features of network nodes through an autoencoder In representation, topic-based network representation learning is obtained. The method of the present invention combines the characteristics of a large-scale information network and a deep learning algorithm, and proceeds from the structure of the information network and the information characteristics contained in the nodes to obtain a more effective embedded representation of the network nodes.

Description

technical field [0001] The invention belongs to the technical field of network representation learning, and more specifically relates to a topic-based network representation learning method. Background technique [0002] Much of the information in everyday life is made up of networks, from social networks to the World Wide Web, which provide a ubiquitous way to organize all kinds of real-world information. Due to the advent of the big data era and the development of deep learning related technologies, how to make full use of complex information networks for data analysis has become a hot research topic in the field of data mining and information retrieval. The central idea of ​​network representation learning is to find a mapping function that converts each node in the network into a low-dimensional latent representation. Network representation learning methods aim to learn dense and continuous representations of nodes in low-dimensional spaces, so that noise or redundant i...

AI Technical Summary

Problems solved by technology

[0007] The problem with existing network representation learning methods is that they only refer to the basic topology of the network, ignoring the rich information of the nodes themselves contained in the real network.

In the real network, the edges are often sparse, and it is difficult for the above method to effectively capture more abundant information for network representation learning.

Images