Author: Abdusalam F Ahmed Nwesri
Title: Answering English Queries in Automatically
Transcribed Arabic Speech
Abstract: There are several well-known approaches to
parsing Arabic text in preparation for indexing and
retrieval.
Techniques such as stemming and stopping have been shown to
improve search results on written newswire dispatches, but
few comparisons are available on other data sources.
In this paper, we apply several alternative stemming and
stopping approaches to Arabic text automatically extracted
from the audio soundtrack of news video footage, and
compare these with approaches that rely on machine
translation of the underlying text.
Using the TRECVID video collection and queries, we show
that
normalisation, stopword-removal, and light stemming
increase retrieval precision, but that heavy stemming and
trigrams have a negative effect. We also show that the
choice of machine translation engine plays a major role in
retrieval effectiveness.
Author: Alex Holkner
Title: Inline dispatch caching for dynamic
languages
Abstract: Dynamic object-oriented languages such as Python,
Ruby, Perl and ECMAScript (a.k.a. JavaScript) allow
relatively unconstrained modification to types and objects
at runtime. This flexibility permits rapid application
development at the expense of runtime efficiency. The most
advanced optimisation techniques such as method inlining
and specialisation cannot immediately be applied in these
languages due to the mutable nature of the object types.
Inline dispatch caching, introduced with the Self compiler
and used in virtually every Java virtual machine today, can
enable these optimisations; but only if suitably extended
to deal with mutable types. We introduce such an extension
and discuss our research proposal, including the relation
of our work to Microsoft's recent foray into the dynamic
language area.
Author: Amirreza Aryani Kashani
Title: Entity-based change propagation
analysis
Abstract: An important part of software evolution is change
impact analysis: determining what are the consequences of a
given change. An initial change in a system can propagate
to other parts of the system, and require further
development and software testing. Change propagation
analysis is important in order to evaluate the cost of
maintenance and software evolution plans. There are a
number of change impact analysis approaches in the
literature, but complexity and specific technical
requirements stop enterprise companies from adopting them
as a generic methodology in the software life cycle. We
propose a novel approach for estimating the scope of change
propagation based on logical relationships between system
components. Our approach is based on the hypothesis that a
dependency map for software components can be derived from
domain-level (abstract) analysis. This approach is
understandable and usable by domain experts who have no
expertise in software architecture. Also it is applicable
to environments where design artifacts or maintenance
history are not accessible. In this paper we focus on
details of the first key step including defining dependency
map, underlying concepts and a method for developing them.
Author: Antony Iorio
Title: Improving the performance of rotationally
invariant differential evolution
Abstract: In many optimization problems it is desirable to
have an optimization algorithm which exhibits behaviour
which is independent of the orientation of the search
space. Differential Evolution is a very popular
optimization algorithm which has this characteristic and
can self adapt to the search space. It has been applied
successfully to many real world optimization problems.
Unfortunately, it has some short comings if rotational
invariance is required, because often insufficient
diversity can be maintained in the population in order for
the search to make progress. Here we will present a novel
Differential Evolution algorithm which samples vectors
around better individuals in the population. This approach
has demonstrated improved performance over a canonical
Differential Evolution scheme, as well as competitive
behaviour compared with Simulated Binary Crossover on the
truss topology design problem.
Author: Bin Rong
Title: Making Application Layer Multicast Reliable
is Feasible
Abstract: Application layer multicast (ALM henceforth) was
proposed as a substitute for network layer multicast~(IP
multicast). However, the end users, who take the
responsibility to replicate and forward data in ALM, are
not as stable as routers in IP multicast. Therefore,
reliability has become the major concern in ALM. This paper
presents
a new tree construction algorithm and demonstrates that
making ALM reliable is achievable, even when a single-tree
based multicast structure is used. It exploits the property
that participating users' lifetime follow a Pareto
distribution which has the used better than new~(UBTN)
feature, and dynamically adjusts the multicast tree.
Participating nodes are organized into a hierarchy and the
hierarchy is organized in such a way that it reflects the
relative stability among participating nodes. The proposed
approach
achieves reliability enhancement for ALM by using a very
low overhead and no a~priori knowledge about usersÇ«÷
lifetime is required. A minimum reduction of $50\%$ can be
achieved in terms of service disruption frequency. Detailed
mathematical analysis and simulation results reveal that
making ALM reliable is feasible.
Author: Carlos Alexandre Queiroz Batista da Silva
Title: Distributed Situations Awareness for
Multi-Agent Systems
Abstract: Many frameworks and architecture models used to
develop agent-based applications are based on the ECA
(Event-Condition-Action) paradigm. When an event is
perceived by the ECA engine it verifies if the event
satisfies one of the conditions registered for it, if it
does, an action is triggered. Some frameworks have extended
the ECA model to support situations which are scenarios
demanding reasoning upon a set of events and in some cases
limited by a temporal space. However these frameworks are
not necessarily adequate for Multi-Agent Systems (MAS).
Multi-Agent Systems are loosely coupled entities (agents)
connected by a network. Agents collaborate with each other
to solve problems or deliver application functionality. Our
research intends to explore a Distributed Situations model,
which is generic enough to be used by any Multi-Agent
systems.
Author: Charles Thevathayan
Title: Composition of security Protocols
Abstract: We present a method for composing security
protocols where multiple message elements using different
levels of security properties can be combined in a provably
secure manner. This is done by using compoable security
patterns and constraints that limit possible actions of
valid protocol participants. The security properties
include authentication, data integrity, secrecy and
non-repudiation allowing eCommerce protocols to be created
from specification.
Author: D. N. F. Awang Iskandar, James A. Thom and S. M. M.
Tahaghoghi
Title: Content-based Image Retrieval Using Image
Regions as Query Examples
Abstract: A common approach to content-based image
retrieval is to use example images as queries; images in
the collection that have low-level features similar to the
query examples are returned in response to the query. In
this paper, we explore the use of image regions as query
examples. We compare the retrieval effectiveness of using
whole images, single regions, and multiple regions as
examples. We also compare two approaches for combining
shape features: an equal-weight linear combination, and
classification using machine learning algorithms. We show
that using image regions as query examples leads to higher
effectiveness than using whole images, and that an
equal-weight linear combination of shape
features is simpler and at least as effective as using a
machine learning algorithm.
Author: Dariush Riazati
Title: Drill Across & Visualization of Cubes
with Non-Conformed Dimensions
Abstract: Data analysts would benefit greatly from the
ability to navigate and view combined multidimensional data
from multiple sources, a key requirement of which is the
conformity between their dimensions. The strict
requirements of conformity restrict navigating to related
multidimensional data from unseen or unfamiliar sources.
In this paper we make a distinction between conformed
dimension tables and conformed dimension attributes and
discuss the merits of relaxing the conformity requirement.
We propose extending the navigation operation drill across
to include the non-conformed dimensions and introduce
Nested Pivot Tables as an extension to Pivot Tables to show
how sources that have conformed as well as non-conformed
dimensions can be viewed and analyzed together.
Author: Dhirendra Singh
Title: Collaborative Learning in Multi-agent
Systems
Abstract: Multi-agent systems (MAS) have shown strong
potential in their applicability to dynamic environments.
Learning in the context of MAS is a valuable tool for
creating more adaptive and robust systems. Furthermore, and
intuitively, a group of learning agents that collaborate
have the potential to achieve better system performance
than when learning individually. In this research we
explore the use of collaborative multi-agent learning
(CMAL) in a dynamic system of multiple agents. The MAL
problem is addressed from an agent paradigm perspective
rather than a Machine Learning one. We consider a practical
application for this work within the Distributed Energy
domain. The scenario is based in a domestic household
equipped with a network of agents that control various
electrical appliances, and where energy consumption is
efficiently managed and controlled through the use of
collaborative learning between the agents.
Author: Gayan Wijesinghe
Title: Using Indexed Nested Loops in Genetic
Programming
Abstract: Genetic programming is used to automatically
evolve programs to solve problems specified by a human
user. Loops, in general, are rarely used in genetic
programming due to the problems of representing them in the
tree structures of the programs, detecting and avoiding
infinite loops, ensuring that they are logically and
structurally valid through their operation and fitness
measurement. In some of our previous work, we have shown
how to successfully use a restricted form of loops, with no
nesting and indexing, for object classification. In this
work, we present a way to implement indexed nested loops in
genetic programming and discuss the how to avoid the
previously mentioned issues when doing so.
Author: Hendrik Gani
Title: Transparently Adding Adaptation Capabilities
to Existing Java Applications
Abstract: Significant development effort is required to
implement adaptive applications that include capabilities
such as collecting contextual information, making
adaptation decisions, and performing the actual adaptation.
Existing work attempts to add transparency by either
partially or completely shifting adaptation responsibility
to a supporting system (e.g. library or middleware).
However in general there is a trade-off between
transparency and adaptation capability since the more
complex or sophisticated the capability, the more
communication required between the application and
supporting system. As such, an extensible and configurable
source based pre-compiler (mobjexc) was developed in
conjunction with a middleware platform for adaptation
(MobJeX), which provides transparent insertion of
middleware-supported adaptation code into existing Java
applications. Furthermore, transparency issues in the
produced code were addressed in order to accommodate
transparent insertion of the current as well as future
capabilities. This paper provides example scenarios where
non-trivial capabilities such as object mobility and
metrics collection are seamlessly inserted into an ordinary
Java application. In addition, an evaluation of the
effectiveness of the proposed solutions in terms of
reducing development effort is provided.
Author: Iftah Gideoni, Lin Padgham, Sebastian Sardina
Title: Merit Based Choice and Commitment in BDI
Agents
Abstract: We intend to amend a typical
Belief-Desire-Intention (BDI) style agent-oriented
programming language so as to enable agents to natively
select and use the best plan from a set of available plans
for serving a goal. In order to tell which course of action
would eventually lead to a preferred state, an agent should
forecast the outcome of each applicable plan. To that end,
we appeal to Hierarchical Task Network (HTN) planning,
which is already embedded within the BDI language to be
used. Because the work is aimed to be applicable to
resource-bounded systems, we shall explore mitigation
techniques to cope with the potentially intractable
planning computation. In particular, we explore ways in
which plans may be abstracted to the so-called
``macro-actions'' and ways in which the embedded planner
may leverage its BDI context to prioritize computation.
We will define the language semantics, show its native
properties with respect to (preferred) plan selection, and
demonstrate its usage in a multi unmanned-aircraft
(simulated) application.
Though the techniques developed are not expected to lead to
optimal solutions, they are expected to cope with
time-constrained domains. Thus, we expect this work to
contribute to the ability to base real-life applications on
formal BDI languages.
Author: Iman Ibrahim Yusuf
Title: Recovery-Oriented Software Architecture for
High Performance Grids(ROSA-Grids)
Abstract: Solving large and time intensive applications
within much shorter period of time is now possible by
running applications on grid environment. In grids, a huge
number of heterogeneous resources that are possibly in
geographically separated areas coordinate to solve
problems. Hence, applications are expected to run reliably
on such environment. However due to the nature, number and
location of resources involved, and the complexity of the
mechanisms needed to integrate these resources into a
seamless utility; the possibility of failure is not only
higher but also inevitable in the environment.
There are exiting fault tolerance approaches for grids. The
approaches are either application specific, restricted to
single domain as opposed to grid definition or limited to
only detect failure. We propose a recovery-oriented
computing (ROC) to grid systems as a fault tolerance
mechanism. Since failure is not a problem rather
unavoidable matter in computing environment, we concentrate
on decreasing the time to recover from possible failure.
In this paper, we survey the combination of ROC with grid
applications. We also analyse how having information about
the architecture of the application benefits us in
developing a better recovery-oriented fault tolerance
mechanism. At the end, we highlight the problems of
existing fault tolerance approaches.
Author: Iman S. H. Suyoto, Alexandra L. Uitdenbogerd, and
Falk Scholer
Title: Effective retrieval of polyphonic audio with
polyphonic symbolic queries
Abstract: Accurately finding audio recordings in response
to symbolic queries is one of the key challenges in the
field of music information retrieval. Pitch is one of the
main features of music; in this paper we propose and
evaluate approaches for using pitch information in
polyphonic symbolic queries to retrieve full tracks of
audio recordings. The audio data is first converted into
symbolic data, using an automated transcription process.
This is a noisy process, adding up to three times as many
notes to the transcription than are actually present.
Nevertheless, recordings can be accurately retrieved by
manually-constructed queries (either in full or truncated)
using the longest common subsequence algorithm (and a
sliding window if the queries are truncated). Precision at
1 of about 80% was achieved, and around 85% of queries
return correct answers in the top 10 from a collection of
1808 recordings. Truncated queries are as effective as
untruncated queries for retrieving correct answers in the
first rank position. Thus, the burden on users is reduced
as they only need to produce a small fraction of a song as
a query.
Author: Jennifer Sandercock
Title: Somatic Marker Agent Development: A
Methodology for Creating Many Characters that Use Emotions
and Personality to Make Decisions
Abstract: Creating virtual characters who use their
emotions and personality to make decisions can be a
time-consuming and difficult task. Little work has been
done previously to address how to easily build unique
personalities based on behavioural choices for a large
number of characters. The methodology presented here shows
a simple method to automatically generate personalities
using Damasio's Somatic Marker Hypothesis. Characters build
up a personality in the form of which choices are better or
worse based on their soft goals and emotional past
experience of what they felt was ``good'' or ``bad''. The
methodology has been implemented generically and allows a
user to easily use the techniques to include emotions and
personality in their own domain dependent environment. The
process is described in detail so the reader can use these
techniques themselves; either from scratch, or using the
software developed by the authors.
Author: Jonathan Yu
Title: Ontology Evaluation using Wikipedia
categories for browsing.
Abstract: Ontology evaluation is a maturing discipline with
methodologies and measures being developed and proposed.
However, evaluation methods that have been proposed have
not been applied to specific examples. In this paper, we
present the state-of-the-art in ontology evaluation -
current methodologies, criteria and measures, analyse
appropriate valuations that are important to our
application - browsing in Wikipedia, and apply these
evaluations in the context of ontologies with varied
properties. Specifically, we seek to evaluate ontologies
based on categories found in Wikipedia.
Author: Jun Jie Foo
Title: Clustering Near-duplicate Images in Large
Collections
Abstract: Near-duplicate images introduce problems of
redundancy and copyright infringement in large image
collections.
The problem is acute on the web, where appropriation of
images without acknowledgment of source is prevalent.
In this paper, we present an effective clustering approach
for near-duplicate images, using a combination of
techniques
from invariant image local descriptors and an adaptation of
near-duplicate text-document clustering techniques; we
extend our earlier approach of near-duplicate image
pairwise identification for this clustering approach.
We demonstrate that our clustering approach is highly
effective for collections of up to a few hundred thousand
images. We also show --- via experimentation with real
examples --- that our approach presents a viable solution
for clustering near-duplicate images on the Web.
Author: Ken Gardiner
Title: A Framework for the Co-Evolution of Genes,
Proteins and a Genetic Code wi
Abstract: This paper presents an artificial chemistry model
where genotypic and phenotypic strings
react with each other. The model prevents the genome from
directly coding for genotype-phenotype mappings or for
gene-replication enzymes. Experiments demonstrate the
genome can evolve to manipulate reactions of phenotypic
strings in such a
way as to alter the genotype-phenotype mapping, and produce
gene-replication enzymes.
Author: Khanh Hoa Dam and Michael Winikoff
Title: Generation of Repair Plans for Change
Propagation
Abstract: One of the most critical problems in software
maintenance and evolution is propagating changes. Although
many approaches have been proposed, automated change
propagation is still a significant technical challenge in
software
engineering. In this paper we present an agent-oriented
change propagation framework based on fixing
inconsistencies when primary changes are made to
design models. A core piece of the framework is a new
method for generating repair plans from OCL constraints
that restrict these models.
Author: Malith Jayasinghe
Title: A Quality of Service Aware Load Distribution
Policy for Web Server Farms
Abstract: Developing efficient load distribution policies
for web server farms has been challenging due to the
complex workload characteristics of the Internet traffic.
Such policies should take into account the high variance in
service time distribution and high variability in arrival
rates. Although there have been several attempts in the
past to develop efficient load distribution polices, these
have several limitations. In this paper, we propose a
Ç«ˇquality of service aware load distribution policyÇ«÷ for
a distributed server farm called QOS-SITA-E. Our QoS aware
task assignment policy is based on the well known SITA-E.
Our model provides Ç«ˇquality of service differentiationÇ«÷
for different classes of clients. We analysis our load
distribution policy analytically and prove that it can
outperform several existing policies under certain
constraints.
Author: Michael C. Harris
Title: What size should a reusable learning object
be?
Abstract: Digital resources for teaching and learning, or
learning objects, have been made available in specialised
repositories as well as on the World Wide Web. Learning
objects may be single files or complex aggregations of many
media assets. When searching learning objects, system
implementors must decide whether it is appropriate to
return an entire learning object or some part thereof. In
this paper we explore the issue of what is an appropriate
resource to return to the user in the context of an actual
query against a repository of learning objects.
Author: Mikhail Perepletchikov
Title: Cohesion Metrics for Predicting
Maintainability of Service-Oriented Software
Abstract: Although Service-Oriented Computing (SOC) is a
promising paradigm for developing enterprise software
systems, existing research mostly assumes the existence of
black box services with little attention given to the
structural characteristics of the implementing software,
potentially resulting in poor system maintainability.
Whilst there has been some
preliminary work examining coupling in a service-oriented
context, there has to date been no such work on the
structural property of cohesion. Consequently, this paper
extends existing notions of cohesion in OO and procedural
design in order to account for the unique characteristics
of SOC, allowing the derivation of assumptions linking
cohesion to the maintainability
of service-oriented software. From these assumptions, a set
of metrics are derived to quantify the degree of cohesion
of service-oriented design constructs. Such design level
metrics are valuable because they allow the prediction of
maintainability early in the SDLC.
Author: Nalaka Dilshan Gooneratne
Title: Matching Strictly Dependent Global
Constraints for Composite Web Services
Abstract: Web service discovery requires matching
techniques for comparing and selecting web service
descriptions based on user constraints. Semantic-based
approaches achieve higher recall than other approaches
(such as syntax-based approaches), because they employ
ontological reasoning mechanisms to match syntactically
heterogeneous descriptions. However, existing
semantic-based approaches are not scalable as they perform
an exhaustive search to locate composite services that
conform to global constraints. This paper proposes a
semantic-based matching technique that locates composite
services. It relates attributes of services to a common
attribute to ensure that they have the same scope. This
enables the assigned values to be compared and evaluated
against a given global constraint. Conforming composite
services are located in polynomial time with a
three-dimensional data structure that indexes services
based on their types, attributes and the assigned values.
Simulation results indicate that the proposed approach
achieves higher recall than syntax-based approaches and is
more scalable than existing semantic-based approaches.
Author: Nicholas Richard May
Title: Improving the availability of composite
services in service-oriented systems
Abstract: Service-Oriented Architecture (SOA) is a style of
software organization that models components as services
and promotes the reuse and sharing of services between
enterprizes. However, SOA does not provide guarantees on
the quality of the services provided, which hinders the
adoption of SOA in business critical systems.
Our project aims to address one of these quality issues by
improving the availability of services through the run-time
optimization of service composition.
In this paper we discuss the goals of the project and the
rationale for the approach selected.
Author: P C Thevathayan
Title: ECommerce Protocols through Pattern
Composition
Abstract: Creating and verifying security protocols remains
a slow process limiting the number of eCommerce
applications. In this paper we propose an approach where
protocols are composed from existing patterns for eCommerce
properties such as authentication, non-repudiation, recency
and data integrity. This approach allows protocols that are
provably secure against known attacks to be created,
avoiding the need for verification.
Author: Paul McIntosh
Title: X3D-UML: Towards Validation of 3D State
Diagrams
Abstract: 3D abstraction techniques for modelling software
systems are not common in the area of software development.
The reasons for the lack of 3D abstraction are unclear and
strong evidence to support or refute the use of 3D, as
applied to software engineering, does not exist. Part of
the reason for the lack of evidence is the difficulty in
producing 3D visualisations and then measuring the benefits
in a way that is not clouded by how the 3D abstraction has
been implemented. This paper builds on previous research,
where a 3D abstraction technique (coined X3D-UML) that is
based on the Unified Modelling Language has shown that UML
classes, already familiar to many software engineers, can
be rendered easily in a 3D space. The initial research went
so far as to mitigate the risk that the X3D, the Web3D
technology used, was not capable of large scale 3D UML
abstraction. This paper presents how this technology is to
be applied to gain strong evidence as to the benefit in use
of 3D for state diagrams.
Author: Pengfei Han
Title: Reducing Overfitting in Predicting
Intrinsically Unstructured Proteins
Abstract: Intrinsically unstructured or disordered proteins
are proteins that lack fixed 3-D structure globally or
contain long disordered regions. Predicting disordered
regions has attracted significant research recently. In
developing a decision tree-based disordered region
predictor, we note that many previous predictors applying
20 amino acid compositions as training parameter tend to
overfit the data. In this paper we propose to alleviate
overfitting in
prediction of intrinsically unstructured proteins by
reducing input parameters. We also compare this approach
with the random forest model, which is inherently tolerant
to overfitting. Our experiments suggest that by reducing 20
amino acid compositions into 4 groups according to amino
acid property can reduce the overfitting in decision tree
model. Alternative approach, ensemble-learning technique
like random forest is naively more tolerant to this kind of
overfitting and can be a promising candidate in disordered
region prediction.
Author: Peter Somerville, Sandra Uitdenbogerd
Title: Note-Based Segmentation and Hierarchy in the
Classification of Digital Musical Instruments
Abstract: The ability to automatically identify the musical
instruments occurring in a recorded piece of music has
important uses for various music-related applications. This
paper examines the case of instrument classification where
the raw data consists of musical phrases performed on
digital instruments from eight instrument families. We
compare the use of extracted features from a continuous
sample of approximately one second, to the use of a
systematic segmentation of the audio on note boundaries and
using multiple, aligned note samples as input to
classifiers. The accuracy of the segmented approach was
greater than the one of the unsegmented approach. The best
method was using a two-tiered hierarchical method which
performed slightly better than the single-tiered flat
approach. The best performing instrument category was
woodwind, with an accuracy of 94% for the segmented
approach, using the Bayesian network classifier.
Distinguishing different types of pianos was difficult for
all classifiers, with the segmented approach yielding an
accuracy of 56%. For humans, broadly similar results were
found, in that pianos were difficult to distinguish, along
with woodwind and solo string instruments. However there
was no symmetry between human comparisons of identical
instruments and different instruments, with half of the
broad instrument categories having widely different
accuracies for the two cases.
Author: Sarvnaz Karimi
Title: Improving Transliteration Effectiveness by
Systems Integration
Abstract: Machine transliteration, the process of rendering
a word from a source language to a target language with
preserved pronunciation, has been studied for a variety of
language pairs. Typically, each of these studies suggest
specific source-target character alignment and word
segmentation methods to form their translation rules.
However, the specific segmentation approach of a
transliteration model can cause errors for some inputs,
leading to incorrect transliteration. In this paper, we
investigate combining the outputs of multiple
transliteration methods. Evaluating the proposed scheme for
Persian-English and English-Persian transliteration, we
observe either improvement in performance in comparison to
using the output of individual systems.
Author: Shahrul Badariah Mat Sah
Title: Animation of photomosaic
composition
Abstract: Photomosaic is a new form of mosaic using
pictures as tiles
to produce the bigger image. Previous researches had
focused
more on producing the final picture in a form of static
display. Although the final composition is an artistic
piece, animating the process of composing it will create
interesting effect on viewers. This paper enhanced the work
of Ciesielski et al.'s idea of animating the process of
generating colour photomosaic through the use of
Evolutionary Computing methods.
Author: Shuhaida Mohamed Shuhidan
Title: Debugging: A constructivist approach to
learning programming
Abstract: The search for effective methods of learning
programming is far from over. This study proposes a
methodology for a novice programmer to learn programming
through debugging. Our study aims to explore debugging to
learning programming, as it has the ability to promote code
reading and tracing, and to generate excitement
during the problem solving process. The constructivist
approaches encourage the learner to construct or build
their own mental model of the knowledge and processes
involved, rather than forcing information to be learnt by
rote, or managed in a top-down manner by the instructor.
The study will look into studentsÇ«÷ learning approaches
and how such learning fits in with constructivism. We
believe that the proposed method will contribute towards a
viable mental model of learning programming and also
enhance the learnersÇ«÷ abilities to construct knowledge by
themselves. The expected learning outcomes are not only
limited to identifying bugs, novice programmers may also
extract design knowledge from the given code, modify code
in order to solve slightly different problems, and develop
mental models of program organization and flow. This
improves critical thinking and problem solving skills, and
also instills in novices the ability to distinguish between
good and bad programs. The study will also investigate the
errors made by novices by identifying and categorizing
their flaws during programming. The approaches to this
study are experimentation via an action research
methodology which incorporates questionnaires and voluntary
focus group participation from first-time programmers.
Author: Simon Alan Duff
Title: Formal Proactive Maintenance Goals
Abstract: In intelligent agent systems, goals are an
important concept, and take a variety of forms. One such
case is are maintenance goals, that in contrast to
achievement goals, define a state that an agent should
continue to keep true, rather than a final state to bring
about. However, agent systems have used maintenance goals
as triggers to goals or plans, rather than being used in
any form of goal deliberation. Our earlier work addressed
this and introduced a notion of proactive maintenance goal,
which enabled an agent to deliberate on all of its goals,
thus producing more rational and intelligent behaviour. The
agent, thru anticipating the effects of its actions, can
avoid violating its maintenance conditions.
This work extends our work in proactive maintenance goals
with the development of formal semantics of maintenance
goals. In contrast to other efforts, our work ensures that
there is a defined declarative link between maintenance
goal and preventative and recovery action.
Author: Stefan Bird, Xiaodong Li
Title: Using Regression to Improve Local
Convergence
Abstract: Traditionally Evolutionary Algorithms (EAs)
choose candidate solutions based on their individual
fitnesses, usually without directly looking for patterns in
the fitness landscape discovered. These patterns often
contain useful information that could be used to guide the
EA to the optimum. While an EA is able to quickly locate
the general area of a peak, it can take a considerable
amount of time to refine the solution to accurately reflect
its true location.
We present a new technique that can be used with most EAs.
A surface is fitted to the previously-found points using a
least squares regression. By calculating the highest point
of this surface we can guide the EA to the likely location
of the optimum, vastly improving the convergence speed.
This technique is tested on Moving Peaks, a commonly used
dynamic test function generator. It was able to
significantly outperform the current state of the art
algorithm.
Author: Steven Burrows
Title: Source Code Authorship Attribution
Abstract: Plagiarism and copyright infringement are major
problems in academic and corporate environments. Existing
solutions for detecting infringements in structured text
such as source code are restricted to textual similarity
comparisons of two pieces of work. In this paper, we
examine authorship attribution as a means for tackling
plagiarism detection.
Given several samples of work from several authors, we
attempt to correctly identify the author of work presented
as a query. On a collection of 1,640 documents written by
100 authors, we show that we can attribute authorship in up
to 67% of cases. This work can be a valuable additional
indicator for the more difficult plagiarism investigations.
Author: Sunidhi Bhalla
Title: Efficient information flow control approach
for web services
Abstract: Web service is a new service-oriented computing
paradigm which poses the unique security challenges due to
its inherent heterogeneity, multi-domain characteristic and
highly dynamic nature. The key challenges in web service
security are access control and information flow control.
Information Flow Control is a generalization of access
control because it controls information after its release,
when access control no longer
applies. Most of the previous work done on access control
and information flow control makes no distinction between
global access control (controlling access of composite web
services) and local access control (controlling access
within a single web service). In this paper we propose the
possible solutions to develop a new integrated security
approach that will deal with large scale distributed
applications of web services by considering global access
control and information flow control
Author: W R M U K Wickramasinghe
Title: User-preference multi-objective optimization
using particle swarms
Abstract: The increase of the number of dimensions or
objectives in a multi-objective optimization problem makes
the search space large that traditional methods of finding
solutions become unfruitful. This is especially the case
when the number of objectives is greater than three. The
introduction of user preferences to the problem is an
efficient search strategy for large search-spaces. When a
user can define the regions in the search-space where
solutions are interesting, the algorithm can concentrate
only in those areas. Here we introduce a user-preference
based particle swarm optimization approach, which is used
to find solutions in multi-objective optimization problems.
This is a novel and efficient approach to find solutions in
multi-objective problem instances which have very large
search spaces.
Author: W. A. Vidura Gamini Abhaya
Title: DNS based Efficient and Scalable load
balancing for Internet Traffic
Abstract: Clustering of servers either locally or
geographically distributed has become a common solution for
popular web sites and content delivery companies. Most load
balancing
solutions at the geographically distributed level are based
on the Domain Name System (DNS). Although DNS provides
great support in its functionality for such situations, its
inherent architecture creates limitations in these
solutions. Many of the current request dispatching policies
are based on primitive measures of parameters such as
server load and domain popularity. Furthermore, the
architecture of the Internet adds to the complexity of
these problems. This research aims to address these
limitations and come up with a more efficient and scalable
solution for DNS based load balancing in geographically
distributed web servers.
Author: Yi Wang
Title: Estimate Singular Point Rotation using
Analytical Models
Abstract: Singular points are important features of
ïngerprints. They play a critical role in ïngerprint
registration and hence many ïngerprint applications. In
this paper, we propose a new approach to estimate the
singular point rotation based on ridge orientation
modeling. The estimation exploits the analytical features
enabled by the orientation model of our previous work. The
accuracy of the proposed estimation approach is assessed
with a ïngerprint rotation database. The performance is
reported with mean squared errors (MSE) and ROC curves.
Author: Yohannes Tsegay, Andrew Turpin, Justin Zobel
Title: Dynamic Index Pruning for Effective
Caching
Abstract: To provide fast query evaluation, search engines
make use of an inverted list stored on disk, caching lists
in RAM wherever possible to reduce query evaluation times.
Dynamic pruning schemes, such as pruning impact-ordered
inverted lists, reduce the amount of each inverted list
processed during query evaluation. Despite only a small
portion of lists being processed when dynamic pruning is
used, current systems still store the entire inverted list
in the cache. In this paper we study the effects of caching
only the pieces of the inverted lists that are actually
used to answer a query when dynamic pruning is used,
leaving the other part of the list on disk. We examine an
LRU cache model, and two other recently proposed models
that have richer eviction policies. We introduce a new
dynamic pruning scheme for impact-ordered inverted lists
that further reduces the volume of lists processed without
degrading answer list quality.
Using two large Web collections and their corresponding
query logs we show that, using an LRU cache, our new
pruning scheme reduces the number of disk accesses during
query processing time by 7-15% over the state-of-the-art
impact-ordered baseline, without reducing answer quality.
Surprisingly, however, we discover that using our new
pruning
scheme makes little difference to disk traffic when the
more sophisticated caching schemes are employed.
Author: Zhiyong Zhang
Title: Automated Unit Testing for Agent
Systems
Abstract: Although agent technology is gaining world wide
popularity, a hindrance to its uptake is the lack of proper
testing mechanisms for agent based systems. While many
traditional software testing methods can be generalized to
agent systems, there are many aspects that are different
and which require an understanding of the underlying agent
paradigm. In this paper we present certain aspects of a
testing framework that we have developed for agent based
systems. The testing framework is a model based approach
using the design models
of the Prometheus agent development methodology. In this
paper we focus on unit testing and identify the appropriate
units, present mechanisms for generating suitable test
cases and for determining the order in which the units are
to be tested, present a brief overview of the unit testing
process and an example. Although we use the design
artifacts from Prometheus the approach is suitable for any
plan and event based agent system.
Author: Qinying Xu
Title: Evaluation of a system for evolving families
of images
Abstract: This research continues the previous work which
described an image generation system called IMAGENE. In
this paper a new texture function called RMod is proposed,
which results in new styles of repeating patterns. We also
build a new system based on IMAGENE called parent system,
which is good at creating families of images. Each family
contains images with similar features, such as shapes or
colours. These images are attractive to artists who work in
ïelds like web design, games and video animation. The
focus of this paper is
on the evaluation of these two new methods. An experiment
of comparing these methods is designed and the analysis of
the results is presented.