Seminar: Dominik Slezak, " Two Case Studies of Rough Set Approaches to Data Mining ...," Oct. 26, 2:30 pm, RIC 208 (Expired)


The Department of Computer Science is pleased to announce a seminar by our former faculty member, Dr. Dominik Slezak.

Title: Two Case Studies of Rough Set Approaches to Data Mining and DataProcessing: Ensembles of Reduct-based Classifiers and Infobright's RDBMS Engine Internals

Speaker: Dr. Dominik Slezak, University of Warsaw and Infobright, Inc.

Date: October 26
Time: 2:30 - 4:00 p.m.
Room: RIC 208 (Research and Innovation Centre)

Abstract:

The theory of rough sets provides clear mathematical and algorithmic
foundations for handling incompleteness and uncertainty in massive amounts
of data. Rough set methods are often utilized in data mining and knowledge
discovery in order to induce various types of decision and classification
models. On the one hand, there are a number of approaches to feature
selection, which refers to the notion of a decision reduct developed within
the theory of rough sets for the purpose of describing irreducible subsets
of features determining decisions at roughly the same level as all
attributes. On the other hand, there are approximated versions of
computational models known from data mining and machine learning, such as
rough clustering, rough support vector machines, or rough neural networks.

In this talk, we refer to both above trends in rough set research and
applications. With regard to the latter one, we show how rough set paradigms
of computing with approximations can be used to scale standard calculations
over huge volumes of data. As an example, we consider Infobrights analytical
RDBMS technology based on hybridization of the principles of columnar stores
and rough computing. We also show some commercial applications of this
database product in the areas of analyzing fast-growing machine-generated
data sets. With regard to the former out of the above-mentioned trends, we
report several extensions of decision reducts aimed at the analysis of
different types of data, for different purposes. In particular, we introduce
reduct ensembles: the families of diversified subsets of features, which
create the basis for pairwise complementary local classifiers in a
classifier ensemble.

In this way, we attempt to show that modern rough set techniques can be
successfully utilized in both scientific and commercial projects, in such
mainstream areas as, e.g., databases and machine learning.


Breaking News


Recent News


Do You Have News for the Department of Computer Science?

  • Please send a plain text version of your posting, which can include URL links, to deptsec@cs.uregina.ca
RSS
To Top of Page