Spring and Summer Computer Science Seminars: Tuesday, June 23 and Thursday, June 25, 2015 (Expired)

You are invited to attend seminar presentations by Computer Science graduate students that will take place during two sessions: on Tuesday, June 23 beginning at 12:45 and Thursday, June 25 beginning at noon.

See below for the schedule.  The seminars will take place in Classroom Building 435.

 

++++++++ Tuesday, June 23, 2015 ++++++++    

                                               

12:45     Jordan Ubbens

 

Boosting and Other Ensemble Methods

 

Ensemble methods are a class of meta-algorithms which improves on a weak classifier to provide strong predictions. This is done by aggregating a set of weak hypotheses. These methods have seen success in many classification problems over their long history and are generally robust to problems associated with overfitting. This seminar will give an overview of ensemble learning in general and two popular ensemble methods in particular, bagging and boosting.

Examples of real-life data sets will be shown where ensemble learning effectively improves classification performance.

               

13:30     Maha El Meseery

 

Geo-Coordinated Parallel Coordinates (GCPC):

A Visual Analytics Approach for Exploring Environmental Data

 

The large number of environmental problems faced by society in recent years has driven researchers to collect and study massive amounts of environmental data. Such environmental datasets are often large and heterogeneous in nature, with complex temporal and geospatial relations. Visual analytics combines multiple visualizations, intuitive interaction, and automated analysis to support reasoning and understanding about massive amounts of data. In this research, a visual analytics approach called Geo-Coordinated Parallel Coordinates (GCPC) is proposed as a method for exploring within environmental data. Multiple coordinated visualizations are used to illustrate the heterogeneous high dimensional, geospatial, and temporal aspects of the data, allowing the user to explore and understand the complex relations in the data.

               

14:15     Radhika Gopi                   

 

Mobile Twitter Search

 

One of the biggest challenges faced by users of social media websites like Twitter and Facebook is content-based exploration of posted messages.

Querying and searching in these websites have made the user put substantial effort by scanning through the entire information. Mobile Twitter Search is a mobile-based search and data mining application that provides users with the ability to explore a large collection of user generated data (tweets) posted on Twitter. This application is focussed on interactive search and exploration supporting the recognition of relevant topics to the search goal, rather than requiring that these be recalled. It fetches the major topics discussed in the textual collection and displays them to the user facilitating further analysis. The features include topic extraction, visual comparison, and exploration of topics and their tweets using a timeline.

 

15:00     Kenneth Emeka Odoh

 

Design and Implementation of a News Dashboard

 

The volume of news generated on the Internet is increasing faster than our ability to read it, thereby necessitating the need for better interfaces.

The list-based representation of existing news interfaces has limitation for news streams. Thus the users miss out on important old news as more recent news is available. This has motivated the design of a news dashboard, following the principles of information visualization and visual analytics.

The goal is to allow users to see and understand the relationship between entities at a glance. The news dataset has been preprocessed to identify topics and named entities. This information is used not only to organize the visual display, but also to automatically identify related, trending, and

breaking news.

 

15:45     Stamatios Katsaganis

 

Character Animation

 

Realistic character animation is an essential component of 3D virtual reality applications ranging from movies to interactive training simulations to video games. In this presentation we will describe skeletal animation, the most popular technique used to represent animated characters in a virtual 3D environment. We will also introduce motion capture, which is the standard method of acquiring motion data that can be used to drive character motion. We will present motion blending as a method for generating new motions from a given set of motion capture data and conclude with an overview of the motion graphs, which is an underlying data structure used by motion blending techniques.

 

++++++++ Thursday, June 25, 2015 ++++++++

 

12:00     Junlong Mo

 

The Correlation Between Evaluation Methods and Bibliometrics

 

A study calculated the citation counts received by each researcher in each of 38 university psychology departments in the United Kingdom. They calculated these counts across each staff and correlated the averages with the Research Assessment Exercise (RAE) ratings awarded to the same department. The correlations were extremely high, which demonstrates RAE process and citation counting measure broadly the same thing. Therefore, they concluded citation counting is more effective and transparent than the system and gives similar results. On the other hand, another study wondered whether there is a correlation between National Institutes of Health (NIH) funding data and bibliometrics. They conclude bibliometric indices are higher for those with NIH funding compared to those without, but only the contemporary h-index was shown to be predictive of NIH funding.

 

12:45     Omar Alali

 

Introduction to Join Tree Propagation Algorithms

 

In recent decades, a few architectures exist in the uncertain reasoning research for definite computation of marginals of multivariate discrete probability distributions. One of the first architectures was proposed by Lauritzen and Spiegelhalter. Hugin and Shafer-Shenoy architectures are alternatives to compute the marginal.  In this seminar, we give an introduction to each of the Lauritzen-Spiegelhalter (LS), Hugin, and Shafer-Shenoy (SS) architectures from different perspectives such as the graphical structure for message propagation, and the message passing scheme in each architecture, including some comparison results between those architectures in term of storage efficiency and computational efficiency.

 

13:30     Baabu Aravind Vellaian Selvarajan

 

Pattern Recognition

 

It is natural that we should seek to design and build machines that can recognize patterns.  Patterns are defined by human beings.  Recognition is thereby related to finding concepts and the art of naming.  Pattern recognition is the act of taking in raw data and making an action based on the "categoryā€¯

of the pattern.  It's a very general technology useful for the automatic detection and classification of patterns in data.  It is the ability to generalize from observations.  The basic steps in pattern recognition, representation and generalization, will be discussed with some examples.

 

14:15     Hadeel Bin Amer

 

Probabilistic Approaches to Card Sorting Analysis

 

A card sorting study of 25 participants using 356 facial photos was conducted. 

Each participant was asked to judge similarity between photos by making an unconstrained number of piles containing similar photos.  Is it possible to gain any insight from this data? A variety of approaches will be presented to better understand whether certain photos, and the judgements about them, may be more important in distinguishing between different strategies that people may use to judge similarity of facial photographs.

 

15:00     Shu Wang

 

Topic: Formulations of Shadowed Sets and Three-way Approximations of Fuzzy Sets

 

Shadowed sets, proposed by Pedrycz, are an example of three-way approximations of fuzzy sets. A fundamental issue in constructing such three-way approximations is the interpretation and determination of a pair of thresholds on the unit interval [0, 1]. In this talk, we review the background of the three-way decisions and fuzzy sets. We introduce a general framework for determining the thresholds.

Within the framework, we critically review existing studies and results and present new formulations according to three principles: uncertainty invariance; minimum distance; and least cost. Finally, we introduce a more general formulation based on the notion the semantic distance, which leads to new results and additional insights.

 

15:45     Xuegang Wang

 

Real-Time Monitoring of Online Auctions Frauds

 

Currently, along with more and more people participating online auctions and more or less experiencing auctions frauds, monitoring fallacious activities becomes critical. Many works have been done regarding to off-line monitoring.

However, it is too late to react to the deceiver after a transaction is completed.

My research is related to real-time monitoring frauds (i.e., shill bidding), which can prevent frauds happening in time. We have proposed a general monitoring framework in our previous work, which divided monitoring steps threefold: detecting frauds, reacting to frauds, and updating clusters of each bidder. In this seminar, I will present shill bidding strategies, fraud pattern metrics, and monitoring algorithms.

Future work will also be presented.


Breaking News


Recent News


Do You Have News for the Department of Computer Science?

  • Please send a plain text version of your posting, which can include URL links, to deptsec@cs.uregina.ca
RSS
To Top of Page