Graph Layout Based on Network Embedding and Improved Dimensionality Reduction

Efficient layout of complex network remains a challenging problem. Based on network embedding and an improved dimensionality reduction technique, we propose a new graph layout method in this work. First, the network embedding method is used to generate the low dimensional dense vector representation of nodes, called embedding, which captures a given node's one-order and high-order neighbourhood node information. And then an improved tsNET approach is proposed to project this embedding to 2D space resulting in a good graph layout. Finally, we evaluate our method comparing with state-of-the-art methods by measuring the layout quality, visual inspection and analyzing the computational complexity. The experiment demonstrates the superiority of our method, and it can produce excellent layouts.