| |
|
 |
Location: 2231 EB-II, Centennial Campus, Raleigh, NC
Email: p-hsiao@ncsu.edu (remove dash)
Tel: 919-513-4663 |
My name is Ping-Lin Hsiao. I go
by
Joe Hsiao.
I am a currently a 3rd-year Computer Science Ph.D Student working in
Knowledge Discovery Lab (KDL) with Chris Healey at NCSU.
My research
interests are
visualilzation
and computer graphics. Here is my resume.
Research
Document Clustering (pdf)
Document clustering is an approach to organize unstructured text
information into meaningful groups. It can be applied to documents in a
database to improve performance in information retrieval, or it can be
used to organize query results from the web or other types of large,
heterogeneous text collections. In this report, we describe a new
clustering algorithm to categorize and spatially cluster text
documents. We employ TF-IDF and term co-occurrence to measure
document-to-document similarities. Next, a modified minimum spanning
tree algorithm is used to cluster similar documents. Finally, we apply
multidimensional scaling on the clusters to represent them as spatial
clusters on a 2D plane. The system is tested with sets of articles
generated by an Internet search engine for certain topic areas. Our
result shows that the system is capable of distinguishing different
topics and producing recognizable and informative clusters.
Master Thesis
 Visualization for
Combinatorial
Auctions (pdf)
We propose a new 2D scheme for concisely visualizing combinatorial
datasets. The visualization
displays concentric rings composed of arcs, with each base element
subset mapped to a
single arc. Equal sized subsets are placed on a common ring. The
outermost ring contains subsets
of size one. Interior rings contain larger subsets. The rings are
positioned to try to overlap
common base elements as much as possible. This allows viewers to search
a local region of
the visualization to study the behavior of a given base element (i.e.,
a given item offered within
the combinatorial auction).
Additional visual features, including motion, color, and texture are
applied to represent auction attributes like the identity of a bidder,
which bids win in a
particular stage of the
auction, and so on. Our visualizations provide viewers with an
efficient and effective way to
observe how an auction progresses.
Project
 Flow
Visualization (exe)
The application takes as input a grid of vector data, and maps those vectors to a grid on the screen. Each coordinate within the grid has two properties: direction and speed, which are inherited from the input vector. The grid is invisible and works like an underlying magnetic field. We chooses a set of seed places on the screen. Each seed place spits out a sequence of bitmaps constantly. These bitmaps travel to different directions with varied speeds, reflecting on which vector the bitmap is on at the current moment. If a bitmap moves onto a new vector, its direction and speed would be changed by the new vector. With enough bitmaps, their movement will reveal the flows of the whole grid of underlying vectors. All bitmaps have a same life cycle. Bitmaps become transparent when they approach the end of life, and disappear at the end. This avoids the screen to be overwhelmed by too many bitmaps. Users can also load their own bitmaps.
|
