genetic algorithm

https://segmentfault.com/a/1190000004155021
http://blog.sina.com.cn/s/blog_4cde15140100gi2u.html
http://blog.csdn.net/v_JULY_v/article/details/6132775

http://arxiv.org/pdf/cond-mat/0501368.pdf
https://www.researchgate.net/profile/Clara_Pizzuti/publication/220739952_Community_detection_in_social_networks_with_genetic_algorithms/links/542ea6a50cf277d58e8ed068.pdf
https://arxiv.org/pdf/cond-mat/0604419.pdf

https://www.researchgate.net/profile/Clara_Pizzuti/publication/221417662_A_Multi-objective_Genetic_Algorithm_for_Community_Detection_in_Networks/links/542ea6a10cf29bbc126f39d9.pdf

http://see.xidian.edu.cn/iiip/mggong/down/PHYSA2012Gong.pdf
http://arxiv.org/pdf/cond-mat/0501368.pdf

https://pdfs.semanticscholar.org/c342/c914dff35d6630c5963a54e3a07adf84f44f.pdf
..........
http://arxiv.org/pdf/cond-mat/0308217.pdf
network modularity

During initial population creation, each node is assigned a random community identifier. However we need a mechanism to give a bias for initial placement of nodes into communities. If two nodes are to be in the same community, they should have connectivity with each other; in the simplest case they might be neighbors. From this assumption, after assigning random community IDs to nodes, we randomly select some nodes and assign their community IDs to all of their neighbors. This bias in the initial population creation improves the convergence of the algorithm and eliminates unnecessary iterations.