Attribute completion and network representation method based on auto-encoder and generative adversarial network

Abstract

The invention discloses an attribute completion and network representation method based on an auto-encoder and a generative adversarial network. The method comprises the following steps: constructing an attribute generator; constructing a graph encoder by using a graph neural network, and taking all attributes and network topology as input and output node representation; constructing a decoder, and using node representation to reconstruct a network topology; constructing an attribute encoder and a structure encoder, and respectively taking attributes and structures as inputs to obtain attribute representation and structure representation; a mutual information estimator is constructed, a raw sample pair is a combination of attribute representation and node representation, and a negative sample pair is a combination of attribute representation and node representation after disturbance; and constructing a discriminator to discriminate the relationship between the attribute and the structure. According to the method, the problem of attribute missing in a real network can be solved, the real attribute can be restored by utilizing the implicit relationship between the attribute and the structure, the node representation process is promoted, and meanwhile, the node representation process is also beneficial to the generation of the real attribute.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to an attribute completion and network representation method based on an autoencoder and a generated confrontation network. Background technique [0002] In complex network analysis, network representation learning has attracted extensive attention because it can well preserve the topological and semantic properties of the original network in the feature space, and the learned representation in the feature space can be applied to many network analysis tasks, Such as community discovery, link prediction, anomaly detection, etc. Graph autoencoders play a key role in network representation learning. A graph autoencoder usually includes a graph encoder and a decoder, where the encoder uses a graph convolutional neural network to effectively fuse network topology and node semantic information (i.e., node attributes) by performing feature propagation and aggregation ...

AI Technical Summary

Problems solved by technology

However, in practical applications, some node attributes in the network are often missing due to privacy issues.

Using traditional data completion methods (such as interpolation methods) to complete attributes and then use encoders for feature extraction often leads to poor results.

At the same time, existing data completion methods ignore the relationship information implicit in node attributes and network topology.

Images