Fig. 2
From: Stochastic block models with multiple continuous attributes
Synthetic Example. We generated a synthetic network with N=200 nodes, K=4 communities and an 8-dimensional multivariate Gaussian for each community. a. A visualization of the adjacency matrix for this network where a black dot indicates an edge. We observe that there is an assortative block structure (blocks on the diagonal), but there are also many edges between communities making the true community structure using only connectivity harder to detect. b. We performed PCA on the N×p attribute array and plotted each of the N nodes in two dimensions. Points are colored by their true community assignments, z. Clustering the nodes according to only connectivity, only attributes, and with the attributed SBM, we quantified the partition accuracy with normalized mutual information, yielding NMI(z,{zconnectivity,zattributes,zattribute sbm})={0.65,0.68,0.83}