Semi-supervised graph representation learning method based on fusion of transfer learning and deep learning and device thereof

Abstract

The invention relates to a semi-supervised graph representation learning method based on fusion of transfer learning and deep learning and a device thereof. The method comprises the following steps: pre-training a graph neural network model through two sub-tasks of a global level and a local level, so that universal representation of input data is learned from unlabeled data; and migrating the pre-trained graph neural network model to the training process of the target task, adding an output layer related to the target task behind the pre-trained graph neural network model, and performing fine tuning on parameters of the pre-trained graph neural network model by using the labeled data to obtain a final graph neural network model. On the basis of saving the manual marking cost, the non-label data and the label data are effectively combined, the generalization ability of the model is improved, the training process of the target task can be simplified, and the purpose of faster convergence is achieved; according to the method, the thought of transfer learning is fully utilized, a large amount of computing resources and computing time can be saved, and the computing efficiency is improved.

Description

technical field [0001] The invention belongs to the field of information technology, and in particular relates to a semi-supervised graph representation learning method and device based on the integration of migration learning and deep learning. Background technique [0002] With the development of Internet technology, large and complex networks in various fields such as social networks, paper citation networks, and transportation networks have penetrated into all aspects of real life in the form of graph data. At the same time, various complex graph mining tasks have emerged. , such as node classification, anomaly detection, link prediction, label recommendation, etc. However, the traditional machine learning feature encoding method, which relies on heuristic models to extract graph structure information, faces unprecedented challenges in dealing with the analysis of these complex scene data. In recent years, the use of deep learning-based graph representation learning met...

AI Technical Summary

Problems solved by technology

However, this type of method only uses the network information of the graph, and the quality of the learned graph representation is not high; 2) Based on the optimization of graph representation learning, that is, by optimizing a clear objective function to learn the low-dimensional representation of the graph, this type of The methods are usually based on domain-related methods, which have poor scalability and need to design objective functions separately for different graph mining tasks; 3) Graph representation learning methods based on deep learning, such as DeepWalk algorithm, Node2vec algorithm, etc., are used in many graph representation learning Remarkable results have been shown in the task, but a large amount of manually labeled data is required

Images