Skip to main content

Table 1 Empirical networks Structural properties of five real-world networks: Escorts, Swarthmore42, Caltech36, Wikipedia, and GitHub

From: Explaining classification performance and bias via network structure and sampling technique

Dataset Escorts Swarth. Caltech Wiki. GitHub
N 16730 1519 701 2132 37700
m 1 1 1 1 1
class role gender gender gender dev
minority escort 2 (m) 1 (f) female 1 (ML)
B 0.40 0.49 0.33 0.15 0.26
E 39044 53726 15464 3143 289003
d 0.0003 0.05 0.06 0.001 0.0004
\(\beta\) 2.87 5.50 4.90 2.87 2.54
H 0.00 0.52 0.54 0.64 0.84
\(N_{fit}\) 14338 208 179 2893 9830
\(m_{fit}\) 2 2 2 2 2
  1. In addition to the properties of interest, we report \(\beta\), the power-law exponent of the degree distribution computed as described in Karimi et al. (2018). \(N_{fit}\) and \(m_{fit}\) represent the number of nodes and minimum degree utilized to generate synthetic networks, respectively