Fig. 5
From: Semisupervised regression in latent structure networks on unknown manifolds
Matrix of scatterplots indicating an underlying low-dimensional structure in the network of 100 Kenyon Cell neurons in larval Drosophila. A directed graph is taken into account where every node represents a neuron. A 6-dimensional adjacency spectral estimate is obtained for every node by augmenting the 3 leading left singular vectors scaled by corresponding singular values, with 3 leading right singular vectors scaled by corresponding singular values. A scatterplot is then obtained for every pair of these 6 components. Since each dimension appears to be approximately related to another dimension via a function, presence of an underlying 1-dimensional structure is assumed