posted Feb 20, 2011 5:55 AM by Ee-Peng Lim
Speaker: Xiaolin Shi, Stanford University Date: Thursday, 24 February 2011, 9:00-10:00am Venue: Meeting Room 4-4 Abstract:
The recent shift of human interaction to the web and online environment presents an unprecedented opportunity to study large-scale social networks and dynamics of human behavior within them. By studying the aggregated behavior of large groups of online users, we are able to uncover interesting behavioral patterns and underlying mechanisms that cannot be observed by examining individual behavior only. In this talk, I will present two related research lines that study aggregated social behavior in information systems. The first is social network analysis. I will give a brief overview of research in social networks as well as an example, which links the behavior that a user will join a group with the features of both the user’s social network and of the group she might join. The second line is the wisdom of crowds. I will show that in the system of an online peer-to-peer loan service, the aggregated dynamic bidding behavior of users reliably predicts the market success of requested loans. Short Bio:
Dr. Xiaolin Shi is a postdoctoral scholar at Stanford University. Her research interests focus on information system and social network analysis, particularly on information dynamics in various types of online social and information networks and how they affect human behavior. Before joining Stanford, she obtained her PhD in Computer Science and Engineering from the University of Michigan in July 2009. Her previous research projects have included characterizing the structural features of networks of online communities, studying the information diffusion patterns and modeling the dynamic process and human behavior in various information sharing networks. She is the recipient of the Douglas Engelbart Best Paper Award at the ACM Hypertext Conference in 2008. More information is available at: http://www.stanford.edu/~shixl/ |
posted Sep 1, 2010 5:44 PM by Ee-Peng Lim
[
updated Feb 20, 2011 5:54 AM
]
18 September 2010 (Saturday) 8:30 am - 12:45 pm Seminar Room 2.4, Level 2
School of Information Systems Singapore Management University 80 Stamford Road Singapore 178902
More detail can be found at:
See article about the workshop at Knowledge@SMU
|
posted Jun 28, 2010 6:20 PM by Ee-Peng Lim
[
updated Jun 28, 2010 6:25 PM
]
Speaker:
| 
Sun-Ki CHAI
Associate Professor
Department of Sociology
University of Hawaii | Date:
Time:
Venue:
| 28 June 2010 (Monday)
10:00 am - 11:30 am
Meeting Room 4.4
School of Information Systems
Singapore Management University
|
Abstract
CLASSIC (Cultural and Link Analysis of Social
Structure of Internet Communities) is a project at the University of Hawai`i to
develop a suite of interrelated software tools that automatically collects and
analyzes a wide variety of network, content, and third-party data on only
communities on the internet. It incorporates a variety of validated social
science theories of culture and action, as well as social network and content
analysis, to identify the “space” on the internet occupied by a particular
virtual community, then to provide information on attitudinal (sentiment,
ideology, emotion), relational (power, influence, access to information), and
behavioral characteristics of the community and its members. Its crawling
algorithms and use of social network theories are the subject of U.S. patent
7499965, “Software Agent for Locating and Analyzing Virtual Communities On The
World Wide Web”.
CLASSIC allows the user to specify the nature of the
community of through by providing one or more seed sites and/or keyword(s), and
then proceeds to locate the boundaries of the community of interest using a
combination of network, content, and other criteria that indicate common
identity. It contains special libraries that automatically recognize
specialized platforms such as forums (BBS) and Twitter, and collects additional
data specific to the platform, a capability that will soon extend to blogs,
social bookmark, and media-sharing sites. Tools are provided for analyzing the
resulting data, and datasets are generated that can easily be imported by
popular statistical, network, and other analysis software. A set of behavioral
prediction tools developed separately is being integrated so that it can be
directly invoked by the crawler, and thus predict real-world behaviors by
members of the virtual community. Advanced users are able to directly specify
the criteria for inclusion in the virtual community, as well as various
technical aspects of the crawl (maximum depth and community size, etc.), but
this is transparent to the beginner. CLASSIC is a research tool that is
designed to optimally provide the data and analysis to will directly address
related research questions that the user needs to answer about online
communities and the terrestrial communities they represent.
About the speaker
Sun-Ki Chai (B.S. Mathematical Sciences, M.S. Computer
Science, Ph.D. Political Science, all Stanford University) is Associate
Professor at the Department of Sociology, University of Hawai`i. His main
theoretical interests are the study of formal models of culture and their
integration with choice-theoretic models of action and network models of
structure, and finally their implementation in software systems. As a
Principal Investigator, he has been awarded grants over the past five years
totaling over $3.5 million, all for interdisciplinary projects that that
integrate social science theory and methods with software development. He is
the author of Choosing an Identity: A General Model of Preference and Belief
Formation (University of Michigan Press, 2001), and co-editor of
Culture and Social Theory (Transaction Publishers, 1998) and
Advances in Social Computing (Springer, 2010). Elsewhere, he has
published papers in journals and edited volumes in disciplines ranging from
sociology, political science, economics, and computer science. He has received
a U.S. patent (#7499965) for development of a specialized web crawler which uses
social science theories and methods to identify and analyze virtual communities. |
posted May 11, 2010 6:33 PM by Ee-Peng Lim
[
updated Feb 20, 2011 5:49 AM
]
Dept. of Computer Science & Engineering
University of Washington
Date: Friday May 21, 2010, 10:00 to 12:00noon
Abstract: Statistical learning handles uncertainty in a robust and principled way. Relational learning (also known as inductive logic programming (ILP)) models domains involving multiple relations. Recent years have seen a surge of interest in the statistical relational learning (SRL) community in combining the two, driven by the realization that many (if not most) applications require both and by the growing maturity of the two fields. Markov logic networks (MLNs) is a statistical relational model that has gained traction within the AI community in recent years because of its robustness to noise and its ability to compactly model complex domains. MLNs combine probability and logic by attaching weights to first-order formulas, and viewing these as templates for features of Markov networks. Learning the structure of an MLN consists of learning both formulas and their weights. To obtain weighted MLN formulas, we could rely on human experts to specify them. However, this approach is error-prone and requires painstaking knowledge engineering. Further, it will not work on domains where there is no human expert. The ideal solution is to automatically learn MLN structure from data. However, this is a challenging task because of its super-exponential search space. We address this problem by presenting a series of algorithms that efficiently and accurately learn MLN structure.
Stanley Kok was born and bred in Singapore. In 1995, he won a scholarship from the National Computer Board of Singapore (now Infocomm Development Authority (IDA)), and gleefully packed his bags to begin his undergraduate education at Brown University, Rhode Island, USA. After spending four wonderful years there (and experiencing enough snow to last him his lifetime), he graduated with honors with a Combined Bachelor of Science (Computer Science) and Bachelor of Arts (Economics) degree. Upon returning to Singapore, he worked as an IT consultant at IDA for several years. Realizing that his first love lay in research, he began his graduate studies in the Computer Science and Engineering department at the University of Washington in Seattle. He received his M.S. in Computer Science in 2005, and a Ph.D. in the spring of 2010. Stanley's research interests are in machine learning, artificial intelligence, and their applications. |
posted May 9, 2010 11:04 PM by Ee-Peng Lim
[
updated May 11, 2010 6:40 PM
]
Abstract:
Graphs are important data for knowledge discovery because a great
deal of global structure emerges from pairwise local affinities.
The standard analysis approach is to construct a matrix from
these local affinities, carry out an eigendecomposition, and then
embed the result in Euclidean space. This, though intricate, is
conceptually simple -- it is based on variation across the global
structure.
In practice, though, spectral techniques require multiple choices. The
theory underlying these choices is not yet well understood, and they also
depend on the goals of the analysis. I will provide a roadmap of
the mapping between problem domains and goals, and algorithm choices,
illustrated with some real-world datasets, for example the social
network of al Qaeda.
Speaker's Biography:
David Skillicorn is a Professor in the School of Computing
at Queen's University in Kingston, Canada. His research focuses on
adversarial knowledge discovery, about which he has published extensively,
including his recent book "Knowledge Discovery for Counterterrorism
and Law Enforcement", Chapman and Hall, 2008.
|
|
