Seminars

The information on this page is not completed and schedule is tentative. Please check back later for more information.

  • Title: Biological Processes as an Inspiration for Generative and Interactive Music System Design: A Survey
    Speaker: Jason Cullimore

    Date: March 14 (M)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Music can be thought of as a human communication system, and it is associated with a wide range of behaviours and cognitive processing strategies. When designing computer software that autonomously creates music, consideration of human behavioural and cognitive systems related to music comprehension and production may serve as an inspiration for the underlying design approach for the software system. In this presentation, I will explore several music generation systems, including the Algorithmic Music Evolution Engine (Hoeberechts, Demopoulos, and Katchabaw, 2007), Norbert Herber's ecologically- and cybernetically-inspired Amergent music systems (2014), and how music psychological theory influenced the design of my own Q-Learning based system for generating chord sequence transitions (2014).

  • Title: Gabor Feature Space in Image Processing
    Speaker: Bingyang Liu

    Date: March 16 (W)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: In image processing, Gabor filters are used for feature extraction, especially for extracting local properties in both spatial and frequency domains. Gabor feature space has been successful used in many applications, for example, representation, processing and segmentation of texture images. The essential of Gabor transform is a special case of short-time Fourier transform. Multiplied by a Gaussian envelop, it can determine the sinusoidal frequency and phase angle for Gabor filters. Based on Gabor feature space, many high level tasks can be performed more efficiently. In this presentation, an overview of the principle of Gabor filter is first given, followed by selected advanced techniques developed recently.

  • Title: A Dynamic Stage-based Fraud Monitoring Framework of Multiple Live Auctionsn
    Speaker: Xuegang Wang

    Date: March 18 (F)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Monitoring progressing auctions for fraudulent bidding activities is becoming crucial in order to detect and stop fraud on time. For this purpose, we introduce a generic stage-based framework to monitor multiple live auctions for in-auction fraud. Creating a stage-based runtime fraud monitoring service is substantially different than has been proposed in the very limited studies on runtime fraud detection. We develop the proposed framework with a dynamic architecture where multiple monitoring agents can be created or deleted with respect to the status of their corresponding auctions (initialized, completed or cancelled). Adopting a dynamic software architecture represents an excellent solution to handle the scalability and real-time performance issues of fraud monitoring systems.

  • Title: On Testing Independencies in Bayesian Networks
    Speaker: Andre dos Santos

    Date: March 21 (M)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Testing independencies in Bayesian networks (BNs) is an fundamental task in probabilistic reasoning. It can reveal the conditional independence relations implied by the directed acyclic graph (DAG) of a BN. One method often utilized for this task is d-separation. Although d-separation has linear time complexity, many have had difficulties in understanding its inner workings. m-Separation is an equivalent method for testing independencies in BNs which coverts the problem into classical separation in undirected graphs. In this seminar we explore the key features of d-separation and m-separation. We show the main advantages and disadvantages of using d-separation and m-separation for testing independencies in BNs.

  • Title: An Introduction to Bayesian Network Inference using Variable Elimination
    Speaker: Jhonatan Oliveira

    Date: March 23 (W)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Bayesian networks (BNs) are a probabilistic graphical model used for reasoning under uncertainty. Queries can be answered in a BN using a process called inference. Newcomers are often introduced to BN inference with a simple algorithm called Variable Elimination (VE). VE can perform inference by summing out variables and multiplying conditional probability tables (CPTs) from the BN. Moreover, VE can save computation under some conditions by detecting and removing certain variables and their respective CPTs which are considered unnecessary for a given query. In this presentation, we give an introduction to BN inference using VE. We also show some of VE's advantages and disadvantages.

  • Title: An Introduction to Fast-Mixed Searching and Related Problems
    Speaker: Yuan Xue

    Date: March 28 (M)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Assume that we want to eliminate a computer virus in a network. We can model this network as a graph, where nodes in the network correspond to vertices and cables between nodes correspond to edges. In order to eliminate the virus and secure the whole network, we will launch a group of searchers into the network. Here comes an optimization question: what is the fewest number of searchers required to perform this task? Graph searching, also called Cops and Robbers games, has become a hot topic dealing with this kind of optimization problems. In this presentation, we give an introduction to the fast-mixed search problem, and establish relations between the fast-mixed search problem and other graph search problems.

  • Title: Overview of Implementations for Behavior Trees and MindSet Editor
    Speaker: Ryan Marcotte

    Date: April 1 (F)
    Time: 1:30pm - 2:20pm
    Place: CL 431

    Abstract: Behavior trees are a popular way of structuring artificial intelligence in games and other virtual reality applications. In a previous seminar, we discussed the theory behind building behavior trees to control game agent behavior and briefly demonstrated MindSet Editor, a graphical editor used for creating and modifying behavior trees. For this seminar, we dive into specific implementation details for behavior trees and how the architecture lends itself well to extension. Then, the flows involved for creating behavior trees in MindSet Editor projects and then exporting those projects to a format consumable by virtual reality applications will be discussed. Behavior trees developed for two simple games will be used to demonstrate the effectiveness of the implementation.



To Top of Page