Dual adaptive learning multi-task multi-view for graph network representation learning.

Graph network analysis, which achieves widely application, is to explore and mine the graph structure data. However, existing graph network analysis methods with graph representation learning technique ignore the correlation between multiple graph network analysis tasks, and they need massive repeated calculation to obtain each graph network analysis results. Or they cannot adaptively balance the relative importance of multiple graph network analysis tasks, that lead to weak model fitting. Besides, most of existing methods ignore multiplex views semantic information and global graph information, which fail to learn robust node embeddings resulting in unsatisfied graph analysis results. To solve these issues, we propose a multi-task multi-view adaptive graph network representation learning model, called M2agl. The highlights of M2agl are as follows: (1) Graph convolutional network with the linear combination of the adjacency matrix and PPMI (positive point-wise mutual information) matrix is utilized as encoder to extract the local and global intra-view graph feature information of the multiplex graph network. Each intra-view graph information of the multiplex graph network can adaptively learn the parameters of graph encoder. (2) We use regularization to capture the interaction information among different graph views, and the importance of different graph views are learned by view attention mechanism for further inter-view graph network fusion. (3) The model is trained oriented by multiple graph network analysis tasks. The relative importance of multiple graph network analysis tasks are adjusted adaptively with the homoscedastic uncertainty. The regularization can be considered as an auxiliary task to further boost the performance. Experiments on real-worlds attributed multiplex graph networks demonstrate the effectiveness of M2agl in comparison with other competing approaches.

Similarity of individual functional brain connectivity patterns formed by music listening quantified with a data-driven approach.

INTRODUCTION: This study aims to explore the similarities in functional connectivity (FC) patterns in individuals when listening to different music genres and, in comparison, to the spoken word, using a novel data-driven approach. Our model and findings can potentially be utilized for evaluating the neurological effects of therapeutic music interventions. MATERIALS AND METHODS: Twelve healthy volunteers listened to seven different sound tracks while undergoing functional magnetic resonance imaging (fMRI) scans: music of the volunteer's choice with positive emotional attachment, two selections of unfamiliar classical music, one classical piece repeated with visual guidance and three spoken language tracks. FC network graphs were created, and selected graph properties were evaluated toward their commonalities across sound tracks. For comparison, FC patterns represented by the graph adjacency matrices were directly compared for high and low BOLD activation during listening. RESULTS: Graph properties averaged across subjects showed similar values for the same sound track compared to different sound tracks (p < 0.003). For high BOLD activation involving most areas in the auditory cortex, FC patterns for the same sound track correlated highly (0.74 ± 0.11), whereas FC patterns for different sound tracks did not (0.09 ± 0.07; p < 6e-5). For low BOLD activation involving additional brain regions, correlation of FC patterns for the sound tracks was still higher (0.43 ± 0.07) than for different sound tracks (0.09 ± 0.05; p < 8e-6). CONCLUSION: Similar music creates similar functional activation and connectivity patterns in the brain of healthy individuals as does listening to the spoken word. Direct comparison of FC patterns yielded higher correlations than indirect comparisons of graph properties derived from corresponding FC networks.