Closed skojaku closed 3 years ago
Results before re-coding. Three algorithms are applied to the political blog network. The vertical axis indicates the normalized mutual information for the metadata (left vs right-leaning of blogs) vs the detected core-periphery paris. The simulations are repeated 30 times and the errorbar shows the variation of the scores attributed to the stochasticity of the algorithms ( calculated with the bootstrapping).
KM_config outperforms modularity maximization and KM_ER in terms of the NMI. KM_config is the most stable and consistently outperforms the other two methods.
Having double-checked my code, I couldn't find any issue. Nevertheless, I re-implement the objective function just in case.
The adjacency matrix of the political blog network with node ordering by the core-periphery pairs detected by KM_config algorithm.
The results are overall the same as that reported in Sadamori Kojaku and Naoki Masuda 2018 New J. Phys. 20 043012
Following issue #15, I added scripts for testing and debugging. More code to come.
Todo