Seminars

  • Title: Fast Searching Problems on Graphs
    Speaker: Dr. Boting Yang (CS)

    Date: Monday, January 30
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Given a graph in which a fugitive hides on vertices or along edges, graph searching problems are usually to find the minimum number of searchers required to capture the fugitive. In this talk, we consider the fast searching, edge searching, node searching, and the relations between them. We also consider some recent work on upper bound, lower bound, NP-completeness and approximation algorithms.

  • Title: Deciphering the biology of Pantoea: a human pathogen in disguise?
    Speaker: Dr. John Stavrinides (Bio)

    Date: Monday, February 6
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: The genus Pantoea respresents a group of bacteria found almost everywhere in the general environment. This bacterial group most often gains attention for its ability to cause diseases in plants; however, Pantoea infections of humans appear to be on the rise, raising questions about the true host range of this species group. To determine the relationship between plant-pathogenic and human-pathogenic isolates, we conducted a phylogenetic analysis of approximately 120 environmental, plant, and clinical strains. Our results indicate that while some species appear to have specialized to either plant or animal hosts, many are mixtures of environmental, plant, and clinical isolates. To determine the host range and pathogenic capabilities of these strains, we have attempted to characterize them by testing their ability to grow in maize (corn) and fruit flies. We show that clinical isolates tend to grow better in fruit flies than in plant hosts, suggesting they are better adapted to insects, and by extension, animals. Surprisingly, many of the clinical isolates retain plant-specific disease genes, indicating that they still may have the ability to infect plants. This suggests that plant-pathogenic isolates of Pantoea may have evolved to associate with animals or insects for dispersal, giving them the capacity to infect humans opportunistically.

  • Title: A graph-theoretic view of computational learning models
    Speaker: Gaojian Fan (CS)

    Date: Wednesday, February 8
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: In many computational learning models, learning algorithms have the goal to identify the target concept from a small number of labeled examples. Different learning models may result in different learning complexities, i.e., the number of examples required in learning. We study different learning models with the focus on graph-theoretic structures of concept classes. The one-inclusion graph is one type of graph has been analyzed in the related literature. As it only provides partial distinguishing information about a concept class, we introduce a new generalized type of graph which provides complete distinguishing information among concepts. We establish connections between those graphs and bounds on learning complexities and give families of concept classes that meet our lower and upper bounds.

  • Title: A Triarchic Theory of Granular Computing
    Speaker: Dr. Yiyu Yao (CS)

    Date: Monday, February 13
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Granular computing is an emerging multidisciplinary study dealing with problem solving and information processing at multiple levels of granularity. A triarchic theory of granular computing is presented and examined. Two fundamental notions of the theory are multilevel granular structures and the granular computing triangle. Granular structures emphasize on multiple levels of understanding and representation of a problem. The triangle is composed of the three pillars of granular computing, namely, the philosophy, the methodology, and the computation. By drawing results from many disciplines, including philosophy, psychology, cognitive science, education, artificial intelligence, computer programming, and many more, the triarchic theory offers a unified framework of granular computing as ways of structured thinking, methods of structured problem solving, and paradigms of structured information processing.

  • Title: Automatically Rigging Characters Using MEL Scripts
    Speaker: Fatimah Masiri (CS)

    Date: Wednesday, February 15
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: In 3D animation, rigging is the process of manually placing skeleton joints inside a character model and attaching them to its surface. Rigging is a tedious task and can deprive beginning animators of the excitement of animation. However, using MEL (Maya Embedded Language) not only simplifies this task, but also allows it to be performed more quickly. This presentation shows how to use MEL to procedurally create a skeleton for a biped character without any intervention from the human animator. It also explains how to use this setup to easily achieve basic animations such as a walk-cycle.

  • Title: Chemistry and the Laboratory for Computational Discovery
    Speaker: Allan East and John Jorgensen

    Date: Monday, February 27
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: The Laboratory for Computational Discovery (LCD) is a collective of CS/Math/Chemistry professors within our Faculty of Science who use Linux supercomputing resources. This talk will discuss supercomputing in scientific research, current LCD resources (hardware and software) and opportunities, and an example of an LCD computational chemistry research project. The two LCD supercomputers, Entropy and Dextrose, provide alternative mechanisms for spreading a problem over multiple processors, and issues such as distributed-memory programming and threading will be discussed. The chemistry project, the electrical conductivity of simulated molten bismuth chloride, is a CPU-intensive project for reasons that will be discussed.

  • Title: Towards a Graphical Depiction of Bayesian Network Inference
    Speaker: Sultan Ahmed (CS)

    Date: Wednesday, February 29
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Bayesian networks (BNs) have been used for managing uncertainty in Artificial intelligence. Variable elimination (VE) and arc reversal (AR) are two basic methods of inference in BNs. It is commonly stated that newcomers have difficulty learning BN inference due to intricate mathematical equations and notations. That is why a graphical method for understanding AR algorithm exists in the literature. In this presentation, we will show our current progress and future plan of depicting BN inference using VE.

  • Title: A New Approach To Face Recognition Based On Generalized Hough Transform and Geodesic Distance Descriptor
    Speaker: Marian Moise (CS)

    Date: Friday, March 2
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Although humans have the ability to recognize faces even from cluttered scenes, implementing such a process on a computer is an overwhelming task because of the multitude of factors that have to be taken into consideration and the associated computing complexity. A new method for recognizing faces from still images will be presented and the results of its evaluation on the Yale database will be analyzed. Face recognition is one of the most prolic research fields and also one of the most demanding. Its complexity is influenced not only by face related attributes such as pose, position, scale, facial expression, accessories and physiognomy changes, but also by environmental factors like illumination, background, occluding objects, and, lastly, camera characteristics. Image descriptors better embed features than do simple pixels, but they have higher memory requirements. Therefore, descriptors are computed for only the most interesting points in an image. A modified version of the Generalized Hough Transform is used to find the image region that best matches the template. Its reference point and hit rate are later used to discriminate between faces. The transform takes into account not only the position of the points, but also the value of their corresponding descriptors which are compared against one another using the Pearson correlation. The most significant advantages of this method are that it requires no training data and that it can be extended to object recognition.

  • Title: TBA
    Speaker: Dr. Yasser Morgan (Eng)

    Date: Monday, March 5
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: TBA

  • Title: A New Formulation of Multi-Category Decision-Theoretic Rough Sets
    Speaker: Bing Zhou (CS)

    Date: Wednesday, March 7
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: As a natural extension to the rough set approximations with two decision classes, I provide a new formulation of multi-category decision-theoretic rough sets. A three-way decision is added to each class which gives the user the flexibility of making a deferred decision. Different misclassification errors are treated separately with the notion of loss functions from Bayesian decision theory. The losses incurred for making deferred and rejective decisions to each class are also considered. The main objective of this approach is to tackle the limitations of the previous related work, and therefore provide a more complete solution to multi-category decision making.

  • Title: Learning Patterns from Helpful Examples
    Speaker: Zeinab Mazadi (CS)

    Date: Friday, March 9
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Patterns are strings containing constant and variable symbols. The language of a pattern is the set of all possible strings generated by the pattern. A typical learning problem is to identify an unknown pattern from "labeled examples", i.e., strings together with positive or negative labels that determine whether or not the strings belong to the language of the unknown pattern. We study the problem of finding a small set of helpful labeled examples in an interactive teaching and learning model. In this model the teacher knows the smallest possible sets of examples that distinguish each specific pattern from others in a target class of patterns.

  • Title: Mining Statistics Canada's Databases For Research Gold
    Speaker: Richard MacLennan (Psychology)

    Date: Monday, March 12
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: In 1996, in partnership with the academic community, Statistics Canada established a new program, called the Data Liberation Initiative (or DLI) to release their data-sets to Canadian post-secondary institutions for research and teaching purposes. As well, in 1998 Statistics Canada embarked on a joint-initiative with SSHRC to facilitate the quantitative training of social scientists in Canada called the Canadian Initiative for Social Statistics (CISS). This second initiative established Data Training Schools (DTS) and Research Data Centres (RDC) to facilitate the use of Statistics Canada data-sets by the academic community. Despite both of these initiatives, there is still a sense that Statistics Canada's databases are being under-utilized in the research community. This presentation will discuss the policies and procedures for accessing Statistics Canada's numerous data-sets for teaching and research purposes. These data-sets present several computing challenges such as working with longitudinal and multidimensional data. This presentation will include a demonstration of the free program Beyond 20/20 to extract two-dimensional data files for statistical analyses, from hyper-dimensional data-sets using the Uniform Crime Reporting (UCR) system. In addition, the implication of employing sampling weights for stratified-random samples employed by Statistics Canada will be discussed. Also the need for employing appropriate procedures for estimating variance components in complex samples will be examined using boot-strapping. Finally, the use of GIS data-sets from Statistics Canada to map data will also be demonstrated.

  • Title: Advanced Interfaces for the Microsoft Kinect System
    Speaker: Natasha Jaques (CS)

    Date: Wednesday, March 14
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Recent developments in computer technology have led to the introduction of devices utilizing new types of user interfaces. One such device is the Microsoft Kinect. The Kinect is able to sense the location and position of a user by projecting an infra-red grid into the area in front of the user. Unlike conventional interfaces, the Kinect can only sense hovering and coarse movements, but cannot detect more fine-grained movements, such as clicking and tapping gestures. Thus, it can be difficult for the user to indicate commands typically associated with these gestures, such as selection and navigation. This presentation will examine various methods for sensing fine-grained movements using an intuitive, quick, and accurate interface for use with the Microsoft Kinect system. In particular, gesture recognition methods informed by depth information will be investigated.

  • Title: Architecture and Algorithms of Multitouch Devices
    Speaker: Joel Rathgaber (CS)

    Date: Wednesday, March 21
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Multi-touch software and technologies need not be limited to smart phone and tablet devices. Do-it-Yourself(DIY) multi-touch devices can be very easy and inexpensive to construct. These devices have research potential in fields such as computer vision, image processing, human-computer interaction and pattern recognition. In this seminar we will explore these DIY multi-touch devices with information on the design and construction using basic tools and materials, the algorithms responsible for enabling rich interaction as well as the protocols used to convey interaction information. Popular development frameworks and software will be investigated along with a discussion on the direction of current research in this field.

  • Title: A Method for Resolving Collisions Between Simulated Spheres in 3D Space
    Speaker: Richard Hamilton (CS)

    Date: Friday, March 23
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: It is often necessary for a computer to know if two objects collide. For example, clicking a button is a collision between a rectangle (the button) and a point (the mouse position). In the game Tetris, the falling tetrominoes stop moving when they collide with other tetrominoes. A physics simulation might need to have gas particles bounce when they collide with each other. In this talk, I will explore a general-purpose method for detecting and resolving collisions between simulated spheres in 3D space. Because spheres can serve as simple approximations of objects from atoms to frictionless cows to spaceships, such a system is useful when speed matters more than accuracy, such as in computer games.

  • Title: TBA
    Speaker: Reza Abbasian (CS)

    Date: Wednesday, March 28
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Constraint Optimization Problems (COPs) are very important real world problems with many applications. A COP is an extension of the Constraint Satisfaction Problem (CSP) in which there is a cost function that should be optimized in addition to satisfying the hard constraints. There are many systematic and non-systematic approaches for solving COPs. In this seminar, I will present my proposed non-systematic approach based on Hierarchical Parallel Genetic Algorithms (HPGAs). I will also present the novel components that I propose within the HPGA. Finally, through an extensive experimental study conducted on several problem instances I will demonstrate the superiority of my proposed approach over other well known evolutionary techniques for solving COPs.

  • Title: Web-based Research Support Systems
    Speaker: Aziz Aljuaied (CS)

    Date: Friday, March 30
    Time: 3:30 p.m.
    Room: CL 408

    Abstract: Web-based Support Systems (WSS) focus on assisting human activities in particular domains based on information and Web technologies. Researchers face many obstacles and therefore need assistance for their research activities. Web-based Research Support Systems (WRSS) provide tools to support some research activities by using Web technology. Research activities can be divided into two levels: the institutional level and the individual level. In this seminar, we will focus on the individual level and also we will study some supporting functionalities in order to build a (WRSS). This system provides some features such as search engine, web pages analyzing, wiki pages, search engine referrers, and forums. The search engine can search one or more sites. It can also search in the site and the researcher has the ability to choose. In addition, any page at the system can host the search engine by having a self-contained search form. The second feature provides tracking for users and pages. It furthermore monitors the links on the system and the files that are downloaded from the pages. The third feature provides WYSIWYG in order to allow the researchers to write articles. In addition, it allows users to edit the work and make comments. The fourth feature extracts the search queries from the search engine in order to know the interesting topics for the visitors. The results of this feature such as the top used keywords and the top pages found. The fifth feature provides a forum for the researchers to exchange knowledge. It gives each user the ability to create his/her forum. It provides forum statistics including number of topics, posts, users, latest user, and currently online users. It displays title of most recent topics in the forum linked to the newest comment in that topic.

To Top of Page