Center for Information Management, Integration and Connectivity
line.jpg (716 bytes)  Home
line.jpg (716
bytes)  About Us
line.jpg (716 bytes)  Contact Us
cimic-noimages.jpg (4076 bytes)

Personnel

Activities

Research

Journals

Publications

Meadowlands Env. Research Inst.

E Commerce

TITLE: (TBA)

Dr. Wei Fan
IBM T.J.Watson Research

February 17, 2006 at 10:30-11:30 AM

 


Machine learning is to fit an accurate inductive model using labelled training data to match a true model. One practical example is to figure out if someone will default a mortgage loan. The major difficulty is that the true model is never known for many real world problems, and any assumptions can be wrong in general. The main stream research of machine learning in the past 20 years has been focusing on rather sophisticated and well-thought approaches to solve problems in classification, regression and probability estimations. Very well known algorithms belonging to the family of sophisticated approaches include Boosting, Bagging, SVM among others.

In this talk, we will introduce Randomized Decision Trees or RDT that can be used effectively for classification, regression and probability estimation problems. The training procedure of RDT incorporates some surprisingly unconventional random factors. However, its accuracy in all three major problems, i.e., classification, regression and probabilitly estimation, are higher and significantly higher those very well known approaches, i.e. Boosting, Bagging, SVM, MLR, and regression trees.

This talk offers the following insights: 1. Introduction of Randomized Decision Trees and its application in classification, regression, probability estimation. Several actual applications within IBM such as equity trading fraud, default payment, information retrieval, storage compoment latency modelling will be included. 2. A fresh and unconventional look at effective machine learning and data mining without strong assumptions.


On the Marketing of Nations: The Shibboleth of Location and Tertiary Education

Ehud Menipaz
The Ira Center for Business, Technology & Society

Friday, February 10th, 2006 at 11:00 am at MEC201

 

One of the most prevalent phenomena of the 21st century is the frequency in which multi national enterprises (MNE) are looking for new location of the next link on the value chain, be it a manufacturing facility or regional R&D headquarters, and the continuous effort by governments, at all levels, to provide the most attractive environment for such endeavours. The presentation deals with two related issues. First, a general framework for multi national enterprises location decision is offered, and the primary factors considered by MNEs when setting up activities through foreign direct investments are presented. This framework utilizes positioning mapping techniques to map manager perceptions of various countries in selected regions and to help policymakers make their countries more appealing for such investments. Second, an inquiry into the relationship between the quality of tertiary educational personnel employed in manufacturing industries and in universities and country competitiveness is discussed. The inquiry uses data derived from the European Union. For both issues, future research and generalizations in scope, geography and industrial sectors are offered.


Mobile Access Control

Adriana Compagnoni
Stevens Institute of Technology (MEC 215)

Friday, February 3rd, 2006 at 11:00 am at MEC201

 

The increasing demands for mobile communications in our society have inspired the academic community to study access control mechanisms in the presence of mobility. In Role-Based Access Control, a given user is assigned a collection of roles (e.g. employee, faculty, student, etc.). In turn, each role is assigned a collection of access privileges. A user gains access to a resource by activating a role which has the necessary privileges. Mobility adds a new dimension to RBAC, since the services available to a given user also depend on the location of the user, agreements between parties, and the technology underlying the connection.
Consider the following example. The University of Wizbrau is equipped with intelligent buildings, and students carry their wireless-enabled laptops with them to class. While in the classroom, students have only limited Internet access and they are not allowed to use e-mail, instant messenger, or visit general web-sites. However, these activities are allowed when done from the student lounge instead. Since the instructor of the course needs a greater access to resources than the students, those activities temporarily disabled to the students are available to the instructor. For example, during a lecture, the instructor may consult her e-mail to address a question raised by a student in an e-mail message.
Traditional RBAC mechanisms associate privileges with roles ignoring locations while our work incorporates location awareness enabling scenarios like the one in our example. This is joint work with Elsa Gunter (UIUC).


[talks in 2004] [talks before 2004]