Fig. 1. Illustration of the DAG and neural networks obtained from a weighted directed graph. The arrows indicate the chemical synaptic connections and their widths indicate relative connection strength, for the types of neurons as indicated in the legend. The original connectome in (A) is turned to the random DAG in (B), by redirecting the edges from high-index nodes to low-index nodes where the node index is randomly assigned as shown in the numbers on the nodes. The constrained DAG in (C) is obtained from (A), by copying the nodes and their positions according to the distance from sensory neurons and then adding the strongest connections first while skipping the cyclic and backward connections. The number on the edges are the rank in weight and dotted arrows indicate the skipped connections. The topology of neural networks (D, E) are obtained by adding source and sink nodes to (B) and (C), before 0-in-degree nodes (input layer) and after 0-out-degree nodes (output layer). The major direction of information flow in (A) is preserved in (E).
© Exp Neurobiol
