Cross-entropy directed embedding of network data

We present a novel approach to embedding data represented by a network into a lowdimensional Euclidean space. Unlike existing methods, the proposed method attempts to minimize an energy function based on the cross-entropy between desirable and embedded node configurations without directly utilizing pairwise distances between nodes. We also propose a natural criterion to effectively evaluate an embedded network layout in terms of how well node connectivities are preserved. Experimental results show that the proposed method provides better layouts than those produced by some of the well-known embedding methods in terms of the proposed criterion. We believe that our method produces a natural embedding of a large-scale network suitable for analyzing by manual browsing in a two- or threedimensional Euclidean space.