Model Comparison

To justify the effectiveness of our proposed CLEVER model, we compared it with the following state-of-arts competitors. Moreover, here we list the brief description and reference of these models and release corresponding sample-code.

PCAKM

K-means is a traditional clustering method to detect communities. To boost the performance and improve efficiency of community detection, we first apply PCA on the original feature spaces and then adopted kmeans on online and offline networks, separately.

Reference

Code

Spectral clustering is one of the most popular modern clustering algorithms, which can detect communities over a single network.

Reference

Code

MMSC

This multi-modal spectral clustering method introduced in Heterogeneous image feature integration via multi-modal spectral clustering is to learn a commonly shared graph Laplacian matrix by unifying different views, where each modal stands for a type of feature from one single view.

Reference

Code

RMKMC

The novel and robust multi-view k-means clustering method proposed in Multi-view k-means clustering on big data is able to uncover the consensus pattern and detect communities across multiple networks.

Reference

Code

CoNMF

It is a co-regularized nonnegative matrix factorization model proposed in Comment-based Multi-View Clustering of Web 2.0 Items by extending NMF for multi-view clustering. It jointly factorizes the multiple matrices through co-regularization.

Reference

Code

CLEVER

We propose a novel co-clustering model for community detection over dual-networks, CLEVER for short. It is capable of reinforcing community detection over online and offline networks by co-regularizing the local and global consistency in a unified model

Reference

Code