Figure 2

E. coli : kernel density estimation for the word nnz D(AUT AUT AU AUT A). Data for two different models are shown: the Kumar model [13] and the Krapivsky-Bianconi [18, 14] model. E. coli is robustly classified as a Kumar network. The Krapivsky-Bianconi model is the runner-up. We here show data for a word that especially favors the Kumar model over the Krapivsky-Bianconi model. The histograms of the word over the training data are shown along with their associated densities calculated from the data by Gaussian kernel density estimation. The densities give the following log-p-values at the word value for the E. coli network: log(p _Kumar ) = -4.22, log(p _KB ) = -12.0.