Fig. 1From: A general deep learning framework for network reconstruction and dynamics learningBasic structure of GGN. Our framework contains two main parts. First, the Adjacency Matrix is generated by the Network Generator via Gumbel softmax sampling; then the adjacency matrix and Xt (node state at time t) are fed to Dynamics Learner to predict the node state in future P time step. The back-propagation process runs back through all the computationsBack to article page