Am Dienstag, dem 28. April 1998, 13.15 -14.45 Uhr hält
Randy Goebel
University of Alberta
Title:
Software development for directory publishing (30 min)
Speaker:
Randy Goebel
University of Alberta
Abstract:
Pre Print Inc. (PPI), with offices in Edmonton and Dresden,
is a software development company that produces software systems for
the directory publishing business. We describe the company, its products,
and its place in the global market, and the exciting evolution of products
they expand to meet the changing demands of the directory publishing
business. A key challenge is managing the evolution of media, as the
world shifts from paper to electronic media. Of paramount importance are
the deployment of modern software development methods, new ideas for
multimedia publishing, and the maintenace of constraint-based layout
and optimization methods.
To help meet some of these challenges, the company has more than doubled
in size in the last year, with continued growth expected, as the demand
for existing and new products increases.
Research talk
Title:
Exploiting hypothetical reasoning for data mining (60 min)
Speaker:
Randy Goebel
University of Alberta
Abstract:
Traditional data mining is commercially attractive because association rules
are easily interpretable by a non-technical audience. To extend data mining research
thus requires a demonstration that more sophisticated learning techniques provide
useful and easily interpretable results in more complex application domains.
This presentation argues that three apparently different learning proble
ms can be usefully
"forced" into a hypothetical reasoning framework. The three problems are the induction of
secondary structure in proteins, the induction of layout rules for two dimensional layout
problems, and the induction of semantic indexing structures for image retrieval.
23.04.98 HG 'Ziegenleder-Saal' 16:00 Uhr
Am Donnerstag, dem 23. April 1998, hält Herr
Dr. Nilo Stolte
Senior Lecturer
Kingston Univertity
London - UK
einen Vortrag zum Thema:
"Robust Hierarchical Voxel Models for Accurate High Quality
Representation and Visualization of 3D Objects"
Representing continuos surfaces accurately into the machine is a big
problem in Computer Science, Computer Graphics and CAD, because all
the well-founded axioms and theorems valid for real numbers are lost
into the machine. The reason is that the machine has limited
resources to be able to represent infinitely refinable surfaces.
Therefore, the representation is usually incomplete and is generally
used for visualization purposes only.
The Robust Hierarchical Voxel Model is a potentially infinitely
refinable model to represent continuous surfaces based on voxels.
Interval arithmetic and implicit surfaces modeling are two powerful
tools used to contruct it. The model is guaranteed to always envelop
the continuous surface it represents. This has dramatically
consequences since a certain numerically accurance can be guaranteed
thus insuring robustness. At the same time it is stored in a
hierarchical structure called octree that allow high voxel space
resolutions with low memory consumption. It also provides a natural
and flexible way to refine the model and to define resolution
independent voxel spaces (where parts of the models can be defined in
higher resolution than others).
These models also enable interactive high quality rendering using Gl
or OpenGL as well as efficient Ray-Tracing.
Impressum:
Herausgeber: Universität Leipzig, Institut für Informatik
Seitenbetreuer: Andreas Zerbst
(e-mail: zerbst@informatik.uni-leipzig.de)
Letzte Änderung:
15.10.98
Status: permanent