Adaptive multi-channel graph convolutional network of joint graph contrast learning

Abstract

The invention discloses a self-adaptive multi-channel graph convolutional network method combined with graph contrast learning. Multi-modal big data is often composed of multiple data of different structural forms, and the data often have the characteristics of complementarity, mutual verification and fusion. How to accurately and efficiently extract complementary information among multi-modal data is a main target of multi-modal research. However, most of methods for performing multi-mode fusion only pay attention to complementary information among multiple modes at present, but often neglect specific information under a single mode. In addition, how to extract abundant and distinctive expressions from multiple modes by using a graph convolutional network is rarely researched at present. Therefore, the invention discloses a self-adaptive multi-channel graph convolutional network method combined with graph contrast learning, a framework takes brain network research as a background, time and space information of a brain network can be mined, and unique and shared features of multiple modes can be effectively fused.

Description

technical field [0001] The invention belongs to the field of machine learning, and in particular relates to graph-based convolutional network and graph comparison learning. Background technique [0002] The brain is the most complex organ in the human biological system, and brain network technology is currently an important method for studying brain diseases, providing a powerful representation of the human-brain interaction model. To conduct non-invasive studies of brain function, a variety of functional brain imaging methods have been employed. In neuroscience, brain networks can often be represented by distinct structural modalities (e.g. diffusion tensor imaging (DTI)) and functional modalities (e.g. resting-state functional magnetic resonance imaging rs-fMRI). These network data represent the complex structure of human brain connections, for example, in rs-fMRI networks, limbic connections represent correlations between brain regions and functional stimuli, while DTI c...

AI Technical Summary

Problems solved by technology

[0006] (2) The embedded expression obtained by using GCN to encode the brain network may not reflect the richness of the original brain network data and structure, especially in the process of encoding, as the number of network layers increases, it is easy to lose information

Images