Skip to main content
Fig. 6 | Applied Network Science

Fig. 6

From: Understanding the limitations of network online learning

Fig. 6

Results of NOL-HTR parameter search on k and ε show benefits and limitations of learning in this setting. Experiments over 56 LFR networks generated using different values of parameters γ, the estimated exponent of the degree distribution (horizontal axis), and μ, the extent of cross-community links, which governs modularity Q (vertical axis). Color of points indicates percent gain of best performing NOL-HTR parameters over the high degree (left) or random degree (right) heuristic methods. NOL-HTR consistently outperforms both heuristics with variation depending on network structure

Back to article page