Ceren Budak

Solving path problems on the GPU

We consider the computation of shortest paths on Graphic Processing Units (GPUs). The blocked recursive elimination strategy we use is applicable to a class of algorithms (such as all-pairs shortest-paths, transitive closure, and LU decomposition without pivoting) having similar data access patterns. Using the all-pairs shortest-paths problem as an example, we uncover potential gains over this class of algorithms. The impressive computational power and memory bandwidth of the GPU make it an attractive platform to run such computationally intensive algorithms. Although improvements over CPU implementations have previously been achieved for those algorithms in terms of raw speed, the utilization of the underlying computational resources was quite low. We implemented a recursively partitioned all-pairs shortest-paths algorithm that harnesses the power of GPUs better than existing implementations. The alternate schedule of path computations allowed us to cast almost all operations into matrix-matrix multiplications on a semiring. Since matrix-matrix multiplication is highly optimized and has a high ratio of computation to communication, our implementation does not suffer from the premature saturation of bandwidth resources as iterative algorithms do. By increasing temporal locality, our implementation runs more than two orders of magnitude faster on an NVIDIA 8800 GPU than on an Opteron. Our work provides evidence that programmers should rethink algorithms instead of directly porting them to GPU.

Structural trend analysis for online social networks

The identification of popular and important topics discussed in social networks is crucial for a better understanding of societal concerns. It is also useful for users to stay on top of trends without having to sift through vast amounts of shared information. Trend detection methods introduced so far have not used the network topology and has thus not been able to distinguish viral topics from topics that are diffused mostly through the news media. To address this gap, we propose two novel structural trend definitions we call coordinated and uncoordinated trends that use friendship information to identify topics that are discussed among clustered and distributed users respectively. Our analyses and experiments show that structural trends are significantly different from traditional trends and provide new insights into the way people share information online. We also propose a sampling technique for structural trend detection and prove that the solution yields in a gain in efficiency and is within an acceptable error bound. Experiments performed on a Twitter data set of 41.7 million nodes and 417 million posts show that even with a sampling rate of 0.005, the average precision is 0.93 for coordinated trends and 1 for uncoordinated trends.

On participation in group chats on Twitter

The success of a group depends on continued participation of its members through time. We study the factors that affect continued user participation in the context of educational Twitter chats. To predict whether a user that attended her first session in a particular Twitter chat group will return to the group, we build 5F Model that captures five different factors: individual initiative, group characteristics, perceived receptivity, linguistic affinity and geographical proximity. Through statistical data analysis of thirty Twitter chats over a two year period as well as a survey study, our work provides many insights about group dynamics in Twitter chats. We show similarities between Twitter chats and traditional groups such as the importance of social inclusion and linguistic similarity while also identifying important distinctions such as the insignificance of geographical proximity. We also show that informational support is more important than emotional support in educational Twitter chats, but this does not reduce the sense of community as suggested in earlier studies.

Big Data in Online Social Networks: User Interaction Analysis to Model User Behavior in Social Networks

With hundreds of millions of users worldwide, social networks provide incredible opportunities for social connection, learning, political and social change, and individual entertainment and enhancement in a multiple contexts. Because many social interactions currently take place in online networks, social scientists have access to unprecedented amounts of information about social interaction. Prior to the advent of such online networks, these investigations required resource-intensive activities such as random trials, surveys, and manual data collection to gather even small data sets. Now, massive amounts of information about social networks and social interactions are recorded. This wealth of big data can allow social scientists to study social interactions on a scale and at a level of detail that has never before been possible. Our goal is to evaluate the value of big data in various social applications and build a framework that models the cost/utility of data. By considering important problems such as Trend Analysis, Opinion Change and User Behavior Analysis during major events in online social networks, we demonstrate the significance of this problem. Furthermore, in each case we present scalable techniques and algorithms that can be used in an online manner. Finally, we propose the big data value evaluation framework that weighs in the cost as well as the value of data to determine capacity modeling in the context of data acquisition.

Diffusion of Information in Social Networks: Is It All Local?

Recent studies on the diffusion of information in social networks have largely focused on models based on the influence of local friends. In this paper, we challenge the generalizability of this approach and revive theories introduced by social scientists in the context of diffusion of innovations to model user behavior. To this end, we study various diffusion models in two different online social networks, Digg and Twitter. We first evaluate the applicability of two representative local influence models and show that the behavior of most social networks users are not captured by these local models. Next, driven by theories introduced in the diffusion of innovations research, we introduce a novel diffusion model called Gaussian Logit Curve Model (GLCM) that models user behavior with respect to the behavior of the general population. Our analysis shows that GLCM captures user behavior significantly better than local models, especially in the context of Digg. Aiming to capture both the local and global signals, we introduce various hybrid models and evaluate them through statistical methods. Our methodology models each user separately, automatically determining which users are driven by their local relations and which users are better defined through adopter categories, therefore capturing the complexity of human behavior.

Information diffusion in social networks: observing and affecting what society cares about

Information diffusion in social networks provide great opportunities for political and social change as well as societal education. Therefore understanding information diffusion in social networks is a critical research goal. This greater understanding can be achieved through data analysis, development of reliable models that can predict outcomes of social processes, and ultimately the creation of applications that can shape the outcome of these processes. In this tutorial, we aim to provide an overview of such recent research based on a wide variety of techniques such as optimization algorithms, data mining, data streams covering a large number of problems such as influence spread maximization, misinformation limitation and study of trends in online social networks.

Data-driven modeling and analysis of online social networks

With hundreds of millions of users worldwide, social networks provide incredible opportunities for social connection, learning, political and social change, and individual entertainment and enhancement in a wide variety of forms. In light of these notable outcomes, understanding information diffusion over online social networks is a critical research goal. Because many social interactions currently take place in online networks, we now have have access to unprecedented amounts of information about social interaction. Prior to the advent of such online networks, investigations about social behavior required resource-intensive activities such as random trials, surveys, and manual data collection to gather even small data sets. Now, massive amounts of information about social networks and social interactions are recorded. This wealth of data can allow us to study social interactions on a scale and at a level of detail that has never before been possible. We present an integrated approach to information diffusion in online social networks focusing on three key problems: (1) Querying and analysis of online social network datasets; (2) Modeling and analysis of social networks; and (3) Analysis of social media and social interactions in the contemporary media environment. The overarching goals are to generate a greater understanding of social interactions in online networks through data analysis, to develop reliable and scalable models that can predict outcomes of these social processes, and ultimately to create applications that can shape the outcome of these processes. We start by developing and refining models of information diffusion based on realworld data sets. We next address the problem of finding influential users in this data-driven framework. It is equally important to identify techniques that can slow or prevent the spread of misinformation, and hence algorithms are explored to address this question. A third interest is the process by which a social group forms opinions about an idea or product, and we therefore describe preliminary approaches to create models that accurately capture the opinion formation process in online social networks. While questions relating to the propagation of a single news item or idea are important, these information campaigns do not exist in isolation. Therefore, our proposed approach also addresses the interplay of the many information diffusion processes that take place simultaneously in a network and the relative importance of different topics or trends over multiple spatial and temporal resolutions.

Mind your Ps and Vs: A perspective on the challenges of big data management and privacy concerns

With large elastic and scalable infrastructures, the Cloud is the ideal storage repository for Big Data applications. Big Data is typically characterized by three Vs: Volume, Variety and Velocity. Supporting these properties raises significant challenges in a cloud setting, including partitioning for scale out; replication across data centers for fault-tolerance; significant latency overheads due to consistency requirements; efficient traversal needs due to high update and velocity; and continuous maintenance in the presence of large variety of data representations. Last but not least, the storage of private data necessitates the ability to efficiently execute queries in a privacy preserving manner (P), without revealing user access patterns. In this paper we highlight these challenges and illustrate sample state of the art solutions.