A controllability identification method based on the deep graph convolutional neural network
A controllability identification method based on the deep graph convolutional neural network(DGCNN)is proposed to address the issues of large matrix operation,high computational complexity and unsuitability for large-scale systems in traditional controllability identification methods for network sys-tem.Under this framework,the aggregation of node and local structural features is first achieved through graph convolutional layers.Then,the aggregated features are fed into the classification pooling layer for sorting and unified output tensor.Finally,a one-dimensional convolutional neural network and fully con-nected layers are added for controllability classification.The results are as follows:The proposed method achieves an AUPR score of over 0.925.Under the same experimental conditions,the calculation time is reduced by 93.3％compared to traditional methods.The results indicate that the method has significant classification performance and outstanding computational advantages,making it have practical application potential in identifying controllability of complex networks.
NETL
NSTL
万方数据
