Identifying Cohesive Subgroups and Their Correspondences in Multiple Related Networks

Abstract

Abstract-Identifying cohesive subgroups in networks, also known as clustering is an active area of research in link mining with many practical applications. However, most of the early work in this area has focused on partitioning a single network or a bipartite graph into clusters/communities. This paper presents a framework that simultaneously clusters nodes from multiple related networks and learns the correspondences between subgroups in different networks. The framework also allows the incorporation of prior information about potential relationships between the subgroups. We have performed extensive experiments on both synthetic and real-life data sets to evaluate the effectiveness of our framework. Our results show superior performance of simultaneous clustering over independent clustering of individual networks.

Enhancing Clustering by Exploiting Complementary Data Modalities in the Medical Domain

“…Note that the third line is a result of the fact that v t+1 minimizes the auxiliary function G , then G ( v t +1 , v t ) ≤ G ( v t , v t ) and F ( v t +1 ) ≤ F ( v t ) as shown in [5]. To conclude the proof of Theorem 1 , we show that the update rule (15) is the update given by (19), i.e.…”

“…Equation (4), however, is the summation of two distance functions. Following the steps in [5] show that minimizing the auxiliary function of the summation is sufficient to decrease the objective function of the sum of distances. The matrix V computed in (15) is used to define the clusters as proposed by [9].…”

“…Unlike previous methods which apply NMF to only one data modality, our proposed method aims to learn from two different modalities simultaneously. Reference [5] proposed to jointly cluster multiple networks using tri non-negative matrix factorization. Their updating rules, however, are different from ours since they minimized the KL-divergence metric in the cost function.…”

“…Their method, however, is semi-supervised and requires user input to allow the algorithm to “learn” the distance metric. Comar, Tan and Jain [5] proposed the joint clustering of multiple social networks to identify cohesive communities characterized by reduced levels of noise. In this paper, we propose BI-NMF that combines information from two complementary data modalities to enhance clustering.…”