|
|
Invited Speaker: Petko Valtchev
Talk: Formal Concept Analysis for Knowledge and Data Discovery: New Challenges
FCA is an approach towards the design of conceptual hierarchies
from observations and therefore is close in spirit to the KDD,
whereby the construction of the concept lattice constitutes the
"mining" phase of the process. FCA offers a framework where classical
KDD problems, such as the compact representation and the visualization
of the mining results are naturally solved. This fact has been
confirmed by a wide range of studies using lattice-based KDD methods
and tools, e.g., for association rule mining, with a steadily and
growing number of practical applications. Nowadays, data mining
faces new challenges where the FCA framework can add value, such
as the dynamicity in datasets, the distribution of data and the
resulting need for integration of partial mining results, the
processing of rich data formats, etc. I illustrate a subset of
these challenges by providing the underlying practical motivation
and then showing how the corresponding difficulties translate into
concrete algorithmic problems within the FCA framework. For each of
the presented problems we then suggest an effective procedure and
study its cost and practical performance. I finally draw some open
issues with FCA-based mining of large datasets.
Biography
Petko Valtchev is an Assistant Professor at the Computer Science and
Operation Research Department of the University of Montreal. He
completed his undergraduate studies (1992) first at Sofia University St
Klyment of Ohrid and then at Vienna University. Petko holds a M.Sc. from
the National Polytechnic Institute of Grenoble (INPG), France, and a
Doctoral degree from J. Fourier University, Grenoble (1999). His Ph.D.
thesis on knowledge discovery for knowledge base construction was
prepared at INRIA. Afterwards, Petko spent a year of Post-doctoral
training at the Department of Computer Science, UQAM, Montreal, Canada,
with a INRIA Post-doctoral Fellowship. He is currently working on
several projects involving formal concept analysis, which receive
financial support from provincial, federal and foreign funding sources,
and is leading the development of the Galicia lattice-manipulation
platform. Petko's research interests include ontology engineering,
knowledge discovery from structured and semi-structured data, algorithms
for lattice construction and maintenance, applications of formal concept
analysis to practical problems from software re-engineering and
visualization, information retrieval, genetics, etc.
ContactE-Mail: valtchev@IRO.UMontreal.CA |
 |
|
