| Authors & Year | Chaotic map used | Ref. no. | Description |
| Liaoruo Wang, Tiancheng Lou, Jie Tang and John E. Hopcroft | Community Kernel Detection, WEBA, GREEDY | [1] | The problem of community kernel detection in large social networks is described as recognizing kernel members and distinguishing the structure of community kernels. They proposed two algorithms for overcoming the issues in one approach. Greedy algorithm which is based on maximum cardinality search. The algorithm efficiently obtains an approximate solution without having a bounded error. Whereas in WEBA they defined and optimized an objective function that is capable of explicitly quantifying the detected community kernels. The algorithm efficiently obtains an approximate solution having a small error bound. |
| Jaewon Yang, Julian McAuley and Jure Leskovec | Communities from Edge Structure and Node Attributes (CESNA) | [7] | While identifying network communities clustering a set of nodes into communities in which a node belongs to multiple communities simultaneously arises a problem. This is due to the fact that nodes in communities possess common properties and attributes having multiple relationships among themselves. They remarked that with the help of two sources of data the clustering task can be performed. They developed an accurate and scalable overlapping community detection method for networks consisting of node attribute information thus providing high performance. They presented Communities from Edge Structure and Node Attributes (CESNA). CESNA is based on a general model for networks with node attributes. Their scheme detected overlapping communities through hard node community memberships. They assumed that communities generate both the network and attributes. Thereby dependency between the network and attributes is possible. For discovering communities they developed a block coordinate ascent method in which all model parameters can be updated in time linear with the number of edges inside the network |
| Jure Leskovec, Kevin J. Lang and Michael W. Mahoney | Graph partitioning, Conductance, Flow based methods. | [8] | They explored various community detection methods in for resolving issues related to the performance and biases of different network community detection algorithms on multiple kinds of networks. They focused on understanding the structural properties of clusters by various methods and then enabling the user to use particular application which would be the most suitable clustering method. They described various classes of empirical evaluations of methods for community detection in networks demonstrating the artifactual properties and systematic biases of community detection objective functions and multiple approximation algorithms. |