Robust graph convolutional neural network method based on space-time sparse learning

Abstract

The invention discloses a robust graph convolutional network method based on space-time sparse learning. According to the method, spatial sparsity is realized on each node through a TopK function, and a time sparsity-based attention mechanism is provided, namely, different weights are allocated to each dimension of a feature space according to different activation frequencies. The invention provides an improved graph convolutional neural network, the original network precision is maintained, the robustness is high, and the anti-interference capability of the model to noise is improved.

Description

technical field [0001] The invention belongs to the field of graph confrontation attack and defense, and in particular relates to a field of improving the robustness of a model in the face of disturbance by using spatio-temporal sparse learning. Background technique [0002] In recent years, the successful application of graph neural networks (GNNs) on various graph-structured data such as social networks, chemical composition structures, and biological gene proteins has attracted more and more attention. However, recent work points out that GNNs are vulnerable to adversarial attacks, which can crash safety-critical GNN applications, such as autonomous driving, medical diagnosis. [0003] The main idea of ​​GNN adversarial attack is to change the topological information of the graph structure or the characteristic information of the nodes to intentionally interfere with the classifier. In terms of adversarial attacks on generated graphs, Dai et al. studied the non-target ev...

AI Technical Summary

Problems solved by technology

However, recent work points out that GNNs are vulnerable to adversarial attacks, which can crash safety-critical GNN applications such as autonomous driving, medical diagnosis

Images