Sixth Annual Student Research Forum

6:45 p.m. Integrating Heterogeneous User Interfaces in Service Oriented Web Applications, Matthias Tilsner

7:15 p.m. BREAK

7:50 p.m. Applying Evolutionary Methods and ARM to Workload Characterization, David Carter

8:10 p.m. Visualizing the Impact of Changes in Software Code, Matthew Follett

8:30 p.m. Chemically Based Topological Indices, Mark Staveley

8:50 p.m. Motion Tracking and Visualization of Wii Remote Controller, Jun Zheng

TUESDAY, MARCH 10^th, 2009, EN-2022

6:00 p.m. Visualization of 3-D Seismic Wave Models, Aaron Maynard

6:20 p.m. The Parallelization of the Lattice Boltzmann Method, Jason Normore

6:40 p.m. The Role of Population Size in Rate of Evolution in Genetic Programming, Ting Hu

7:15 p.m. BREAK

7:50 p.m. Photo Browsing 2.0, Grant Strong

8:10 p.m. Construction of n-regular Register from Safe Registers, Ranjeet Kumar

8:30 p.m. Freeing the Health Information Access Layer with a Medical Context Messaging Framework, Jamie Goodyear

Closing Remarks

* * *

MONDAY, MARCH 9^th, 2009

6:05 p.m. Mark Hatcher, Computer Science Ph.D. student

(Supervisors: Dr. Wolfgang Banzhaf and Dr. Gwoing Yu)

Title: Principled Artificial Chemistry

Abstract: Artificial chemistry draws loose inspiration from real-life chemistry, with a typical system defining some types of molecule, how they collide and the reactions that happen when they collide. What happens in the system over time depends on all three elements. Usually the available set of reactions is pre-determined, and behaviour is fine tuned by manipulating the types and amounts of molecules available and how they collide. Some work has explored the evolution of the reaction set from a large, unrestricted search space, however adopting an ‘anything goes’ approach can lose practical benefit. Conversely, principled artificial chemistry requires that all reactions in the set obey a set of principles, yet still allows evolution of the reactions, the principles or both. This talk will explain what model extensions the principled version allows, how it might be of practical benefit, and the costs involved in its development and use.

* * *

6:25 p.m. S. Mehdi Fatemi B., Computational Science M.Sc. student

(Supervisors: Dr. Gwoing Yu and Dr. Siu O’Young)

Title: On the Multi-layer Behaviours of Hybrid Dynamical Systems

Abstract: “Advances in networked embedded computing and communication devices have fuelled the need for design techniques that can guarantee safety and performance specifications of embedded systems, or systems that involve the integration of discrete logic with the analog physical environment.” Hybrid dynamical systems intrinsically combine non-linear continuous dynamics and discrete logical transitions among a number of different operating modes. This presentation highlights our new insight into different functional levels in the general behaviour of hybrid systems. In addition, we describe our novel multi-layer framework to have a complete theoretical and computational backbone to model and control hybrid systems accurately.

* * *

6:45 p.m. Matthias Tilsner, Computer Science Ph.D. student

(Supervisor: Dr. Adrian Fiech)

Title: Integrating Heterogeneous User Interfaces in Service Oriented Web Applications

Abstract: Service Oriented Architecture (SOA) is a design framework that permits the developer to build complex applications from smaller modules called “services”. When applying SOA to web applications, developers face several challenges. Among others, the heterogeneous user interfaces provided by these services must be adapted to fit into the larger application.

My research discusses approaches to automated adaptation of web application’s layout and design. I review the range of web-based user interface modifications, suggest a technique that allows us the automated execution of the described steps and finally explore transparent integration of the modified user interfaces into a given web page flow (using a proxy server). My approach relies on the proposed meta-language UDLD, which permits application providers to equip their user interfaces with semantic information about included components.

* * *

7:15 p.m. BREAK

7:50 p.m. David Carter, Computer Science student

(Advisors: Dr. Wolfgang Banzhaf and Dr. Orland Hoeber)

Title: Applying Evolutionary Methods and ARM to Workload Characterization

Abstract: While modeling methods such as queuing models are fairly well understood, workload characterization remains a bit of a black art. In simple cases, workloads are described completely by a process. In more typical cases, a single process spans multiple workloads, which may in turn span multiple processes or even multiple servers. Numerical optimization methods are commonly applied to model calibration, but not to workload characterization, primarily due to a lack of accurate transaction information.

The Application Response Measurement (ARM) standard supplies both transaction counts and response times. Combining this with evolutionary numerical optimization methods we are able to solve for workload characterization as well as model calibration.

* * *

8:10 p.m. Matthew Follett, Computer Science M.Sc. student

(Supervisor: Dr. Orland Hoeber)

Title: Visualizing the Impact of Changes in Software Code

Abstract: In software projects, there are often teams of programmers all working on the same code. In any given team, there is a chance of ill-written code being introduced. While syntax problems are easily identifiable by compiling, logical errors are only noticeable during execution tests. In some cases, these tests may not be possible until after the code has been committed to the repository. While some programs use a graphical means of showing when a line of code in a project was last updated, these systems do not recognize that function calls may hold miswritten code written within them. These same functions can be called multiple times, in various program files, and may be difficult to detect under these current state of the art methods. In this presentation, I will show traditional and graphical code tracking systems, and introduce a new concept to visualize code changes, using a node-link diagram representation that illustrates how the changes may have impact on the system as a whole.

* * *

8:30 p.m Mark Staveley, Computer Science Ph.D. student

(Supervisors: Dr. Sharene Bungay and Dr. Ray Poirier)

Title: Chemically Based Topological Indices

Abstract: Chemical structures can be represented in a number of different ways, each with their own strengths and weaknesses. For example, Cartesian coordinate representations capture the 3-dimensional information about the structure but are very difficult to search and browse. Whereas, the Simplified Molecular Input Line Specification (SMILES) supports textual based searching but does not represent any of the 3-dimensional information about the structure. Representation schemes for chemical structures are typically based upon either graph theory or chemical semantics that are then converted into some type of readable or parsable format. Since there are significant shortcomings with these approaches we have developed a representation scheme that uses both the chemical and topological information that is contained within a chemical structure.

This seminar will first discuss issues relating to the representation of chemical information using graph theory and chemical semantics. This will then be followed by a presentation of the theory and related examples of our method for capturing chemical information.

* * *

8:50 p.m. Jun Zheng, Computer Science M.Sc. student

(Supervisor: Dr. Minglun Gong)

Title: Motion Tracking and Visualization of Wii Remote Controller

Abstract: The Wii Remote controller (or Wiimote) is an innovative 3D input device. Compared with traditional gaming controllers which usually support 2D operations, it can track movement in 3D space. The Wiimote uses three accelerometers to measure accelerations along the X, Y, and Z annex. When a user waves the Wiimote in a certain gesture, the accelerometers will record the values of accelerations. By tracking the acceleration data we can acquire the movement of the user. One problem of the Wiimote is the lack of decision when used as a pointing device. To increase the accuracy, a camera is attached in front of the Wiimote. Then, we can apply image processing techniques to the pictures captured by the camera to correct the positioning. The other problem is that the feedbacks of the controllers are acceleration data. It is difficult to image the gestures from these values. We aim to use information visualization techniques to provide more intuitive data based on those elusive feedbacks.

* * *

TUESDAY, MARCH 10^th, 2009

6:00 p.m. Aaron Maynard, Computer Science M.Sc. student

(Supervisor: Dr. Minglun Gong)

Title: Visualization of 3-D Seismic Wave Models

Abstract: Effective visualization of 3-D geological data-volumes is of great interest to professionals in the hydrocarbon extraction industry, as well as earth scientists researching earthquakes and other subsurface phenomena. However, most visualization packages only display static models constructed from seismic field data; they do not show how the energy in the seismic wave interacts with and distorts the model data. Because of the steadily increasing speed and storage capacities of computing devices, visualizing seismic events as 3-D animations has become a possibility, even approaching real-time performance. As a solution, a system is proposed to perform a visualization of a wave equation simulation through a seismic model.

Challenges encountered include choice of suitable hardware, balancing trade-offs between simulation speed and accuracy, and between model size and rendering performance, among others.

* * *

6:20 p.m. Jason Normore, Computational Science M.Sc. student

(Supervisor: Dr. Wolfgang Banzhaf)

Title: The Parallelization of the Lattice Boltzman Method

Abstract: The Lattice Boltzmann Method (LBM) is a relatively new and exciting alternative to the well-established methods of fluid system simulations. It uses a discretized set of particle distribution functions to represent a system. The Lattice Boltzmann method is ideal for a parallel implementation because of its discretization and the nature of the method itself. This talk will consist of a brief introduction to the LBM, followed by an outline of how it can be parallelized and some performance results of the parallelization method. The primary focus of the talk will be the technique used to parallelize the LBM, which is an overlapped domain partitioning technique, customized for the Lattice Boltzmann Method.

* * *

6:40 p.m. Ting Hu, Computer Science Ph.D. student

(Supervisor: Dr. Wolfgang Banzhaf)

Title: The Role of Population Size in Rate of Evolution in Genetic Programming

Abstract: Population size is a critical parameter that affects the performance of an Evolutionary Computation model. A variable population size scheme is considered potentially beneficial to improve the quality of solutions and to accelerate fitness progression. In this work, we discuss the relationship between population size and the rate of evolution in Genetic Programming. We distinguish between the rate of fitness progression and the rate of genetic substitutions, which capture two different aspects of a GP evolutionary process. We suggest a new indicator for population size adjustment during an evolutionary process by measuring the rate of genetic substitutions and effectively accelerate fitness progression. A test with the Mackey-Glass time series prediction verifies our observations.

* * *

7:15 p.m. BREAK

7:50 p.m. Grant Strong, Computer Science M.Sc. student

(Supervisor: Dr. Minglun Gong)

Title: Photo Browsing 2.0

Abstract: This talk will basically be an update on the state of our current research on organizing and browsing large sets of community photos. I will discuss interface improvements like a new static photo display order that addresses the previous weaknesses caused by dynamic k-means clustering, that is the flickering. I will also present the findings from our recent work on the evaluation of different image feature vectors, both color and/or gradient-based, for the application of photo organization. A short demo will be provided at the end for a practical look at how everything has been coming together.

* * *

8:10 p.m. Ranjeet Kumar, Computer Science M.Sc. student

(Supervisor: Dr. K. Vidyasankar)

Title: Construction of n-regular Register from Safe Registers

Abstract: Hardware construction of a regular register being difficult, it is desirable to be able to construct a regular register from safe registers, which can easily be incorporated in the hardware. A safe register is the one which may return any value within its domain in case of overlapping writes. On the other hand, a regular register has the property that it either returns the old value or the new value, in case of overlapping writes with the read operation. 1-regular register has already been constructed using 3 safe registers. 1-regular register implies that the register returns a regular value (old or new value) if at most 1 write is overlapping with the read operation. We have extended this idea to construction of an n-regular register, using safe register, such that it may return a regular value (old or new value) if no more than n writes are overlapping with the read operation i.e. an n-regular register can handle n overlapping writes.

PRESENTATION SCHEDULE