Mixture models for learning low-dimensional roles in high-dimensional data

Somaiya, Manas; Jermaine, Christopher; Ranka, Sanjay

doi:10.1145/1835804.1835919

Cited by 7 publications

(8 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The task of role discovery has been studied in different types of graphs, e.g., social networks [22]. Different approaches have been used for role discovery, including Bayesian frameworks using MCMC sampling algorithm for learning multiple roles of data points [23], semi-supervised semantic role labeling [24], etc. There is another related body of work in role mining, a nice overview of which is given by Molloy et al [25].…”

Section: Related Workmentioning

confidence: 99%

RolX: Role Extraction and Mining in Large Networks

Henderson¹,

Gallagher²,

Eliassi-Rad³

et al. 2011

204

287

View full text Add to dashboard Cite

Abstract-Given a graph, how can we automatically discover roles for nodes? Roles could be, eg., 'bridges', or 'peripherynodes', etc. Roles are compact summaries of a node's behavior that generalize across networks. They enable numerous novel and useful network mining tasks, such as sense-making, searching for similar nodes, and node classification. We propose RolX (Role eXtraction), a scalable (linear in the number of edges), unsupervised learning approach for automatically extracting roles from general network data. We demonstrate the effectiveness of RolX on several network mining tasks, from exploratory data analysis to network transfer learning. Moreover, we compare network role discovery with network community discovery. We highlight fundamental differences between the two (e.g., roles generalize across disconnected networks, communities do not).

show abstract

Section: Related Workmentioning

confidence: 99%

RolX: Role Extraction and Mining in Large Networks

Henderson¹,

Gallagher²,

Eliassi-Rad³

et al. 2011

204

287

View full text Add to dashboard Cite

show abstract

“…However, as the research areas of subspace clustering [15] and multi-view clustering [17] have taught us, for many data collections multiple, differing aspects of the observations are captured. This aspect has already been touched by approaches like [22,11,3] that allow for a mixed membership in different components. Thus, they realize an overlapping clustering and, e.g., allow for a movie to participate in the 'humor' as well as in the 'action' genre [3].…”

Section: Fig 1: Example For the Multi-view Scenariomentioning

confidence: 99%

“…the likelihood of belonging to a cluster), such a principle of soft clustering does not support the idea of generating objects through multiple components as for the multi-view scenario. To overcome this issue, a few models [22,11,3] try to represent such multi-component membership (i.e. overlapping clusters).…”

Section: Related Workmentioning

confidence: 99%

Multi-view clustering using mixture models in subspace projections

Günnemann

Färber

Seidl

2012

Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

Detecting multiple clustering solutions is an emerging research field. While data is often multi-faceted in its very nature, traditional clustering methods are restricted to find just a single grouping. To overcome this limitation, methods aiming at the detection of alternative and multiple clustering solutions have been proposed. In this work, we present a Bayesian framework to tackle the problem of multi-view clustering. We provide multiple generalizations of the data by using multiple mixture models. Each mixture describes a specific view on the data by using a mixture of Beta distributions in subspace projections. Since a mixture summarizes the clusters located in similar subspace projections, each view highlights specific aspects of the data. In addition, our model handles overlapping views, where the mixture components compete against each other in the data generation process. For efficiently learning the distributions, we propose the algorithm MVGen that exploits the ICM principle and uses Bayesian model selection to trade-off the cluster model's complexity against its goodness of fit. With experiments on various real-world data sets, we demonstrate the high potential of MVGen to detect multiple, overlapping clustering views in subspace projections of the data.

show abstract

“…Moghaddam et al [19] provided a systematic method to process the billions of records in this netflow to integrate and aggregate the multidimensional data. The Global Local model mainly uses a generic Bayesian framework called the PrObabilistic Weighted Ensemble of Roles Model, or "POWER" model [24] for short. This framework is a new class of mixture models [7,17] where multiple components can contribute to the generation of a single data point while simultaneously allowing each component to have a varying degree of influence on different data attributes.…”

Section: Introductionmentioning

confidence: 99%

“…It has been shown that it takes 510 hours to learn the model on a data set collected from Reuters news stories (22,429 stories, 21 mixture components, 1,000 word attributes) using a 32-core machine [24]. Running the netflow dataset on the model will be impractical.…”

Section: Introductionmentioning

confidence: 99%

A global local modeling of internet usage in large mobile societies

Almutairi

Somaiya

Moghaddam

et al. 2012

Proceedings of the 7th ACM Workshop on Performance Monitoring and Measurement of Heterogeneous Wireless and Wired Networks

Self Cite

View full text Add to dashboard Cite

Real-world wireless Internet usage data for any user is typically generated via an overlap of many correlations. These correlations could be based on hobbies (e.g. sports-fan), profession (e.g. work e-mail), day-to-day activities (e.g. news, Internet banking), communication (e.g. instant messaging, social networks), etc. The likelihood of appearance of these correlations in usage data may be influenced by the type of location the user is in. Hobbies and communication related web sites would be more likely to be accessed at home, Profession related web sites would usually be accessed at work. Understanding and capturing this generative process that is based on human interests, behavior and location is the key to the design of future mobile networks. We propose a novel Bayesian mixture model called the "Global Local" model based on the "POWER" model that can realistically describe Internet usage and correlations with various locations inside a large mobile society. The "POWER" model is a new class of mixture models where components compete to produce a single data point, this model allows for the discovery of complex overlapping patterns of user's Internet behavior. The "Global Local" model learns a global template of user's Internet behavior patterns using the "POWER" model first, then learns correlations between the templates and locations inside a large mobile society. We design a learning algorithm that can effectively learn the "Global Local" model from Internet usage data, and demonstrate its capabilities using synthetic data. Finally, we analyze a real-world Internet usage data for thouPermission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. sands of users collected via wireless LAN traces and discover many interesting correlations that can be explained very intuitively.

show abstract

Mixture models for learning low-dimensional roles in high-dimensional data

Cited by 7 publications

References 11 publications

RolX: Role Extraction and Mining in Large Networks

RolX: Role Extraction and Mining in Large Networks

Multi-view clustering using mixture models in subspace projections

A global local modeling of internet usage in large mobile societies

Contact Info

Product

Resources

About