Understanding the limitations of network online learning

Applied Network Science

Table 2 Coefficients of a linear regression using the output of our NOL-HTR parameter search

The regressor matrix was made up of the following statistics for each network in the search: global clustering, modularity (Q), and estimated degree exponent (γ). The response variable was the best performing ε. Results suggest that higher modularity predicts higher ε, suggesting that networks with modular structure benefit most from exploration