Arun S. Maiya

Sampling community structure

We propose a novel method, based on concepts from expander graphs, to sample communities in networks. We show that our sampling method, unlike previous techniques, produces subgraphs representative of community structure in the original network. These generated subgraphs may be viewed as stratified samples in that they consist of members from most or all communities in the network. Using samples produced by our method, we show that the problem of community detection may be recast into a case of statistical relational learning. We empirically evaluate our approach against several real-world datasets and demonstrate that our sampling method can effectively be used to infer and approximate community affiliation in the larger network.

Online sampling of high centrality individuals in social networks

In this work, we investigate the use of online or “crawling” algorithms to sample large social networks in order to determine the most influential or important individuals within the network (by varying definitions of network centrality). We describe a novel sampling technique based on concepts from expander graphs. We empirically evaluate this method in addition to other online sampling strategies on several real-world social networks. We find that, by sampling nodes to maximize the expansion of the sample, we are able to approximate the set of most influential individuals across multiple measures of centrality.

Aggression, grooming and group-level cooperation in white-faced capuchins (Cebus capucinus): insights from social networks

The form of animal social systems depends on the nature of agonistic and affiliative interactions. Social network theory provides tools for characterizing social structure that go beyond simple dyadic interactions and consider the group as a whole. We show three groups of capuchin monkeys from Barro Colorado Island, Panama, where there are strong connections between key aspects of aggression, grooming, and proximity networks, and, at least among females, those who incur risk to defend their group have particular “social personalities.” Although there is no significant correlation for any of the network measures between giving and receiving aggression, suggesting that dominance relationships do not follow a simple hierarchy, strong correlations emerge for many measures between the aggression and grooming networks. At the local, but not global, scale, receiving aggression and giving grooming are strongly linked in all groups. Proximity shows no correlation with aggression at either the local or the global scale, suggesting that individuals neither seek out nor avoid aggressors. Yet, grooming has a global but not local connection to proximity. Extensive groomers who tend to direct their efforts at other extensive groomers also spend time in close proximity to many other individuals. These results indicate the important role that prosociality plays in shaping female social relationships. We also show that females who receive the least aggression, and thus pay low costs for group living, are most likely to participate in group defense. No consistent “social personality” traits characterize the males who invest in group defense. Am. J. Primatol. 73:821–833, 2011.

Inferring the Maximum Likelihood Hierarchy in Social Networks

Individuals in social networks are often organized under some hierarchy such as a command structure. In many cases, when this structure is unknown, there is a need to discover hierarchical organization. In this paper, we propose a novel, simple, and flexible method based on maximum likelihood to infer social hierarchy from weighted social networks. We empirically evaluate our method against both simulated and real-world datasets and show that our approach accurately recovers the underlying, latent hierarchy.

The Impact of Structural Changes on Predictions of Diffusion in Networks

In a typical realistic scenario, there exist some past data about the structure of the network which are analyzed with respect to some possibly future spreading process, such as behavior, opinion, disease, or computer malware. How sensitive are the predictions made about spread and spreaders to the changes in the structure of the network? We investigate the answer to this question by considering seven real-world networks that have an explicit timeline and span a range of social interactions, from celebrity sightings to animal movement. For each dataset, we examine the results of the spread analysis with respect to the changes that occur in the network as the time unfolds as well as introduced random perturbations. We show that neither the estimates of the extent of spread for each individual nor the set of the top spreaders are robust to structural changes. Thus, analysis performed on historic data may not be relevant by the time it is acted upon.

Exploratory analysis of highly heterogeneous document collections

We present an effective multifaceted system for exploratory analysis of highly heterogeneous document collections. Our system is based on intelligently tagging individual documents in a purely automated fashion and exploiting these tags in a powerful faceted browsing framework. Tagging strategies employed include both unsupervised and supervised approaches based on machine learning and natural language processing. As one of our key tagging strategies, we introduce the KERA algorithm (Keyword Extraction for Reports and Articles). KERA extracts topic-representative terms from individual documents in a purely unsupervised fashion and is revealed to be significantly more effective than state-of-the-art methods. Finally, we evaluate our system in its ability to help users locate documents pertaining to military critical technologies buried deep in a large heterogeneous sea of information.

Topic similarity networks: Visual analytics for large document sets

We investigate ways in which to improve the interpretability of LDA topic models by better analyzing and visualizing their outputs. We focus on examining what we refer to as topic similarity networks: graphs in which nodes represent latent topics in text collections and links represent similarity among topics. We describe efficient and effective approaches to both building and labeling such networks. Visualizations of topic models based on these networks are shown to be a powerful means of exploring, characterizing, and summarizing large collections of unstructured text documents. They help to “tease out” non-obvious connections among different sets of documents and provide insights into how topics form larger themes. We demonstrate the efficacy and practicality of these approaches through two case studies: 1) NSF grants for basic research spanning a 14 year period and 2) the entire English portion of Wikipedia.

Expansion and decentralized search in complex networks

Borrowing from concepts in expander graphs, we study the expansion properties of real-world, complex networks (e.g., social networks, unstructured peer-to-peer, or P2P networks) and the extent to which these properties can be exploited to understand and address the problem of decentralized search. We first produce samples that concisely capture the overall expansion properties of an entire network, which we collectively refer to as the expansion signature. Using these signatures, we find a correspondence between the magnitude of maximum expansion and the extent to which a network can be efficiently searched. We further find evidence that standard graph-theoretic measures, such as average path length, fail to fully explain the level of “searchability” or ease of information diffusion and dissemination in a network. Finally, we demonstrate that this high expansion can be leveraged to facilitate decentralized search in networks and show that an expansion-based search strategy outperforms typical search methods.

A Framework for Comparing Groups of Documents

We present a general framework for comparing multiple groups of documents. A bipartite graph model is proposed where document groups are represented as one node set and the comparison criteria are represented as the other node set. Using this model, we present basic algorithms to extract insights into similarities and differences among the document groups. Finally, we demonstrate the versatility of our framework through an analysis of NSF funding programs for basic research.