2010 CS Research Expo
Research Presentation Abstracts
Speaker:
Ken Kent
Presentation
Title: Accelerating Java Using Massive Multicore
Systems
Abstract
High-performance, low-cost
processing capabilities are essential to supporting the competitive
implementation of information technology systems. Until
recently, hardware performance enhancements focused on increasing
the clock speed of processors, thereby increasing the number of instructions
executed per unit time. This performance improvement did not require any
changes to software systems. However, processor clock speeds have now reached
their practical limit. To continue to deliver processing improvements, chip
manufacturers have resorted to placing multiple processors on a single chip:
so-called multi-core architectures. Multi-core architecture performance
increases are dependent on design changes in the software systems to distribute
the software execution over multiple processors. Currently, this means the
software must be rewritten to make explicit use of parallel and distributed
computing algorithms.
The objective of this
project is to perform the basic research, applied research and productization necessary to create a set of software tools
and techniques to allow Java programs to run efficiently on massive multi-core
systems, which are rapidly being adopted as the high-performance processors of
choice, even in computer systems of moderate cost (e.g. PS3). The project is
comprised of three components which together form a stand-alone product with
significant market potential for organizations developing Java software. These
components are 1) a software tool or tools to convert existing Java programs
for accelerated massive multi-core performance; 2) a modified Java class
library which supports accelerated multi-core, and can be used with existing
Java programs; and 3) modified Java components which support accelerated
multi-core. Together, these components form a complete solution by providing a
massive multi-core runtime which improves the performance of existing Java
programs, as well as the software tools to further transform Java programs for
accelerated multi-core.
Speaker:
Wei Song
Title: Wireless Networking for Multimedia
and Healthcare Applications
Abstract
The next-generation wireless communications have
been envisioned to be supported by heterogeneous networks using various
wireless access technologies. IEEE 802.11 wireless local area networks (WLANs),
also known as Wi-Fi, perform well to
provide last-mile connectivity by radio links, whereas the 3G cellular networks
are not the only choices to serve as a wireless backbone for hotspots. Another
viable option is WiMAX (worldwide interoperability
for microwave access) based on IEEE 802.16 standards. WiMAX
offers broadband wireless access over long distances with a less expensive and
easier-to-build infrastructure. It is well accepted that next-generation
wireless networks will integrate multiple broadband wireless accesses to
provide “always best connectivity.”
In our research, we focus on the state-of-the-art wireless networking
technologies to integrate the complementary strengths of different networks to
offer multimedia and healthcare applications. In particular, we are
investigating high-quality multimedia service provisioning under diverse
mobility situations including a vehicular environment. Also, we are studying the
interconnection of multiple wireless systems to
timely and seamlessly forward the vital life signals monitored from body
sensors to a remote healthcare center.
Speaker:
Chris Baker
Title:
On the Illustration of Ontology
Reuse
Abstract
The economic benefit of ontology
engineering remains an ongoing point of discussion in the community at large,
in part because the value and reuse of semantic metadata is not well
demonstrated in applications and use cases. In particular illustrations of
the architectural combination and synergy of semantic with other technologies
remain limited. This presentation will review phases of the knowledge
engineering cycle and outline reuse scenarios of ontologies
at different levels of maturity and formalization. In particular ontology reuse
in the following scenarios will be demonstrated: Ontological Natural
Language Processing for knowledge discovery, Biomolecule
Classification and Nomenclature Revision, Semantic Assistant Web Services
for desktop applications. The examples will use ontologies
from Telecom and Life Science domains.
Speaker: Huajie Zhang
Title: Semi-supervised
learning and its applications
Abstract
Machine
learning has been widely used in various applications, such as search engines,
text classifications, e-commerce, etc. Most machine learning techniques need a
set of labelled data. In some applications, however,
it is often expensive or difficult to obtain labelled
data. One the other hand, unlabeled data are often relatively easy to collect.
Semi-supervised learning targets on establishing an accurate model by using
unlabeled and labelled data together. This talk
introduces briefly the general idea, major techniques and applications of
semi-supervised learning.
Speaker: Stuart MacGillivray
Title: I/O-Efficient Data
Structures
Abstract
When
dealing with massive amounts of data, structures and operations may be best
optimized to minimize disk I/Os rather than in-memory operations. Various
fields require storage of large numbers of spatially oriented data points, and
efficient range searches on that data. To that end, we are studying the
d-dimensional grid file as proposed by Nievergelt et
al, and have proposed extensions for expanding its capacity and for adding data
persistence to the structure.
Speaker:
Judy Zhao
Title:
Managing Uncertain Knowledge on the
Fuzzy Semantic Web
Abstract
Uncertainty is an intrinsic
feature of real-world knowledge. Our work is motivated by the limitations
of precise knowledge reasoning. This talk first gives some
background on the semantic web and zooms
into uncertainty-related issues. We then present our approach of
dealing with uncertain knowledge on the fuzzy semantic web. Specifically,
we discuss our uncertain knowledge representation formalism and its
core reasoning algorithm on top of a combination of description logic and
fuzzy logic. Our approach will be illustrated with a travel and
a medical use case.
Speaker:
Duong Dai Doan
Title:
Making Inferences About Family
Genetics
Abstract
Human genomes contain two
copies of each chromosome. Research shows that single chromosomes, called haplotypes, are useful to study complex genetic diseases.
While genomic data, called genotypes,
are abundant and easy to collect, haplotypes are rare
and much more difficult to obtain by a biochemical method. Therefore, a
computational method to infer haplotypes
from genotype data, called
haplotyping, is important. Genotypes can be obtained
from a population group where relationships between members are unknown or from
a multigenerational family pedigree with known relationships between members.
In the absence of recombination
events, haplotypes of members in a pedigree follow
the Mendelian law of inheritance, where the
two haplotypes of a child are transferred from its parents, one haplotype from its father and the other from its mother.
Polynomial time haplotyping algorithms are known for
non-recombinant pedigree data. If recombinations
are allowed, the problem of inferring haplotypes for pedigrees with the minimum number of recombinations is NP-hard, even for general pedigrees with
only two sites or tree pedigrees with multiple sites.
We study the minimum haplotype configuration for general pedigrees, where each
member in a pedigree has only two sites. We show that this NP-hard problem can
be reduced
to the Bipartization by Edge Removal problem. Our problem
can be solved by a fixed-parameter algorithm with a running time of O(2k n2), where n is the number of members and k is the number of recombination events.
We
extend our algorithm to handle
general pedigrees with a small number of sites. Our extended algorithm can correctly
infer haplotypes for a general pedigree in O(2k2m^2n2m3)
time where m
is the number of sites in each member.
Speaker:
Sangwhan Cha
Title:
Middleware Framework for Mobile Application Services
Abstract
Mobile application
services are prone to failures caused by disruption of an active wireless
access network connection due to the user’s movement to other networks or
signal blocking (shadowing). Therefore proper mechanisms for disconnection
tolerant mobile application services are needed. In our research, we
investigate a middleware framework that transparently performs required
functionality for users in order to provide continuous mobile application
services in case of network disruption. Such middleware framework provides
effective disconnection tolerant mobile application services including services
for mobile video, mobile conference, mobile games, etc.
Speaker:
Nicola Bicocchi, Post Doctorate, UNB
Title: Collaborative Organizational Modelling and Simulation for Emergency-Response
Abstract
The main objective of this
research is to provide insights into the complex intra- and
inter-organizational challenges that adversely impact the emergency-response
capabilities of modern countries. Our goal is to model and simulate
organizations involved in this domain. To this end, we specify a collaborative
capability to model and simulate emergency-response organizations using
state-of-the-art modelling and simulation frameworks
and we demonstrate the benefits of our approach using a proof-of-concept
simulation.