|
Invited speakers
Santosh S. Vempala (Georgia Tech)
homepage
Effective Principal Component Analysis
Principal Component Analysis (PCA) is arguably the most widely used technique on
high-dimensional or large data sets. This is despite the fact that for standard
applications (finding nearest neighbors, clustering, learning etc.), it is easy
to build examples on which PCA *fails*. In this lecture, we discuss
some problems where the performance of PCA is provably near-optimal, and no other
method is known to have similar guarantees. The problems include:
(a) unraveling a mixture of unknown Gaussians and
(b) learning a function of an unknown subpsace.
On the way, we describe extensions of standard PCA that are noise-resistant,
affine-invariant and use higher moments, thus addressing some of the
know drawbacks
of standard PCA.
Pavel Zezula (Masaryk University in Brno, Czech Republic)
Future Trends in Similarity Searching
Similarity searching has been a research issue for many years, and searching
has probably become the most important web application today. As the
complexity of data objects grows, it is more and more difficult to reason
about digital objects otherwise than through the similarity. In this
article, we first discuss the concepts of similarity and searching in light
of future perspectives before a concise history of similarity searching
technology is presented. We use the historical knowledge to extend the
trends to future. We analyze the bottlenecks of application development and
discuss perspectives of search computing for future applications. We also
present a model of search technology and its position in computer clouds for
application development. Finally, execution platforms for multi-modal
findability and security for outsourced similarity searching are suggested
as important research challenges.
|