Abstract
Due to the importance of community structure in understanding network and a surge of interest aroused on community detectability, how to improve the community identification performance with pairwise prior information becomes a hot topic. However, most existing semi-supervised community detection algorithms only focus on improving the accuracy but ignore the impacts of priors on speeding detection. Besides, they always require to tune additional parameters and cannot guarantee pairwise constraints. To address these drawbacks, we propose a general, high-speed, effective and parameter-free semi-supervised community detection framework. By constructing the indivisible super-nodes according to the connected subgraph of the must-link constraints and by forming the weighted super-edge based on network topology and cannot-link constraints, our new framework transforms the original network into an equivalent but much smaller Super-Network. Super-Network perfectly ensures the must-link constraints and effectively encodes cannot-link constraints. Furthermore, the time complexity of super-network construction process is linear in the original network size, which makes it efficient. Meanwhile, since the constructed super-network is much smaller than the original one, any existing community detection algorithm is much faster when using our framework. Besides, the overall process will not introduce any additional parameters, making it more practical.
Highlights
The main idea is to construct a super-network based on the network topology (Fig. 1(a)) and pairwise prior information (Fig. 1(b)), which is equivalent to the original network topology with smaller size and tight formulation and preserves the must-link pairwise prior information
To demonstrate its high accuracy and speed, we take the framework from Zhang et al.[12] as baseline for comparison. Both our approach (Super-Network) and Zhang’s (ModTop) modify the network topology according to the pairwise constraints and can be readily used in many existing community detection methods
Normalized Mutual Information (NMI)[20] and run time given in seconds are used to measure the accuracy and efficiency, respectively
Summary
Zhang et al modify the network adjacency matrix according to the pairwise prior information and apply existing community detection algorithms to the modified network[11, 12]. Different from passive techniques, active semi-supervised community detection techniques, i.e., semi-supervised community detection based on active learning, assume that pairwise prior information is not given in advance and design the algorithm to select pairs of nodes critical for performance improvement, for human labeling[15, 16]. Yang’s unified semi-supervised framework introduces a parameter balancing the impact of the topology information and priors to maximize the performance improvement[14]. All of these drawbacks limit the application of these methods for problems.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
