APICS 2000 Conference Program

TIME TABLE

FRIDAY, OCTOBER 20, 2000
Time	Activity	Location
12:30 - 6:00	Computer Science Competition	CS Floors 2 & 3
2:00 - 5:00	Mathematics Competition	Chase Building
2:00 - 4:30	APICS CS Committee Meeting	CS Dean's Office
2:30 - 4:30	APICS Math/Stat Committee Meeting	CS Room 311
5:00 -	Registration	CS Lobby
5:00 -	Book Exhibition	CS Lobby
6:00 - 7:00	Pizza Party	CS Atrium
	(for participants in competition)
7:00	Official Welcome	CS Auditorium
7:10 - 8:00	Blundon Lecture: J. Lambek	CS Auditorium
8:00 - 10:00	Wine and Cheese	CS Basement
SATURDAY, OCTOBER 21, 2000
8:45 -	Registration	CS Lobby
8:45 -	Book Exhibition	CS Lobby
9:00 - 9:50	Invited Lecture I: M. Gentleman	CS Auditorium
9:50 - 10:10	Coffee Break	CS Basement
10:10 - 12:10	Math/Stat Student Paper Session	CS Seminar I
	Appl Math & Numer Anal Session	CS Seminar II
	CS Student/Contributed Papers	CS Auditorium
12:10 - 2:00	Lunch and Awards	CS Basement
2:00 - 2:50	Invited Lecture II: L. Bretthorst	CS Auditorium
2:50 - 3:10	Coffee Break	CS Basement
3:10 - 4:00	Invited Lecture III: J. Slonim	CS Auditorium
3:10 - 5:40	Math/Stat Contributed Papers	CS Seminar I
	Appl Math & Numer Anal Session	CS Seminar II
4:00 - 5:00	CS Contributed Papers	CS Auditorium
SUNDAY, OCTOBER 22, 2000
9:30 - 11:30	AARMS Session: Bayesian Statistics	Chase Colloquium Rm
9:00 - 1:00	AARMS Session: Category Theory	Chase Seminar Room
10:30 -	AARMS Business Meeting	Chase Conf. Room

Conference Schedule

FRIDAY, OCTOBER 20, 2000
BLUNDON LECTURE
Location: Computer Science Auditorium
Chair: Robert Paré

Time	Speaker and Title
7:00pm - 8:10pm	Joachim Lambek, McGill University
	Mathematics and the Mind

SATURDAY, OCTOBER 21, 2000
INVITED LECTURE I
Location: CS Auditorium; Chair: Arthur Sedgwick

Time	Speaker and Title
9:00am - 9:50am	Morven Gentleman, Dalhousie University
	Relative Gaps Between Highly Composite Numbers

INVITED LECTURE II
Location: CS Auditorium; Chair: George Gabor

Time	Speaker and Title
2:00pm - 2:50pm	Larry Bretthorst, Washington University
	Bayesian Spectrum Analysis and Parameter Estimation

INVITED LECTURE III
Location: CS Auditorium; Chair: Arthur Sedgwick

Time	Speaker and Title
3:10pm - 4:00pm	Jacob Slonim, Dalhousie University
	Impact of Technology on Society

SATURDAY, OCTOBER 21, 2000
MATH/STAT STUDENT PAPER SESSION
Location: Computer Science Seminar Room I
Chair: Keith Johnson

Time	Speaker and Title
10:10am - 10:25am	Cedric Davies, Saint Mary's University
	Tutte Connectivity in Graphs
10:25am - 10:40am	Alasdair Graham, Memorial University of Newfoundland
	Area and Perimeter Approximation of Circles within
	a Square
10:40am - 10:55am	Colin Macdonald, Acadia University
	The Predicted Sequential Regularization Method of
	Differential-Algebraic Equations
10:55am - 11:10am	Mary Maclachlan, Acadia University
	Modelling Electrical Activity in the Human Heart
11:10am - 11:25am	Sarah Mathews, Memorial University of Newfoundland
	Cyclically Indecomposable Two-fold Triple Systems
	from Skolem Sequences
11:25am - 11:40am	Christopher Pollock, Mount Allison University
	Extended Langford Sequences
11:40am - 11:55am	Ian Rutherford, Mount Allison University
	A Generalization to the Quantum Binomial Theorem
11:55am - 12:10pm	Robin Swain, Memorial University of Newfoundland
	8-Cycle Decompositions of the Cartesian Product of
	Two Complete Graphs

SATURDAY, OCTOBER 21, 2000
APPL MATH & NUMER ANAL SESSION
Location: Computer Science Seminar Room II
Organizers: Pat Keast(Dal), Ray Spiteri(Acadia) and Shigui Ruan(Dal)

Time	Speaker and Title
10:10am - 10:35am	John Stockie, University of New Brunswick
	Multicomponent Gas Transport in Fuel Cell Electrodes
10:40am - 11:05am	Rebecca Culshaw, Dalhousie University
	A Delay-Differential Equation Model of Cell-to-Cell
	Spread of HIV
11:10am -11:35am	Jeff McNally, University of New Brunswick at Saint John
	A Stacking Algorithm for Solving Tri-diagonal Symmetric
	Toeplitz Systems of Linear Equations
11:40am - 12:05pm	Jing Zhang, University of New Brunswick at Saint John
	Parallel Projection Methods on a Beowulf Cluster
3:10pm - 3:35pm	Sarita S. Nemani, Univ. of New Brunswick at Saint John
	Solving Banded Linear Systems
3:40pm - 4:05pm	Andy Foster, Memorial University of Newfoundland
	Strong Periodic Forcing in a Predator-prey Model
4:10pm - 4:35pm	James Watmough, University of New Brunswick
	A Simple SIS Epidemic Model with a Backward Bifurcation
4:40pm - 5:05pm	Xingfu Zou, Memorial University of Newfoundland
	An Embedding Technique for Monotonicity in a Neural
	Network with Delay
5:10pm - 5:35pm	Hermann Brunner, Memorial University of Newfoundland
	The Numerical Solution of Volterra
	Integro-differential Equations:
	a Survey of Current Research and Open Problems

SATURDAY, OCTOBER 21, 2000
CS STUDENT/CONTRIBUTED PAPER SESSION
Location: Computer Science Auditorium
Chair: Arthur Sedgwick

Time	Speaker and Title
10:10am - 10:25am	Fauzi M Ali, University of New Brunswick
	Toward Faceted Classification Scheme of
	Software Design Models
10:25am - 10:40am	Paul Mountford and P. Lingras, Saint Mary's University
	Time-delay Neural Networks designed using Genetic
	Algorithms
10:40am - 10:55am	Colin Cherry and A. Trudel, Acadia University
	An SQL-based Self-modifying Intranet Search Engine
10:55am - 11:10am	Denis Riordan, Dalhousie University
	A Virtual Live Learning Room for First Year
11:10am - 11:40am	Weichang Du, University of New Brunswick
	Towards a Logical Basis for Modeling and Querying
	Distributed Multidimensional Information Systems
11:40am - 12:10pm	Jamie Blustein, Dalhousie University
	An Experiment in Evaluating Automatically
	Generated Hypertext
4:00pm - 4:30pm	Nur Zincir-Heywood, Dalhousie University
	Corporate IT Architectures
4:30pm - 5:00pm	Malcolm Heywood, Dalhousie University
	Problem Solving from Nature

SATURDAY, OCTOBER 21, 2000
MATH/STAT CONTRIBUTED PAPER SESSION
Location: Computer Science Seminar I
Chair: Jeannette Janssen

Time	Speaker and Title
3:10pm - 3:30pm	Joe Apaloo, St. Francis Xavier University
	Dynamic Stability of Evolutionary Games
3:30pm - 3:50pm	Robert Dawson, Saint Mary's University
	New Tilings of the Sphere with Congruent Triangles
3:50pm - 4:10pm	Megan Dewar, Dalhousie University
	An Interval Colouring Approach to the Computer Test
	Scheduling Problem
4:10pm -4:30pm	M. Kotchetov, Memorial University of Newfoundland
	Polycharacters of Cocommutative Hopf Algebras
4:30pm - 4:50pm	Dan Kucerovsky, University of New Brunswick
	Introduction to the Algebraic Topology of C*-Extensions
4:50pm - 5:10pm	Weijiu Liu, Dalhousie University
	Stabilization and Controllability for
	the Transmission Wave Equation
5:10pm - 5:30pm	Eric Marchand, University of New Brunswick
	On a Particular Sum of Dependent Bernoulli and
	its Relationship to a Matching Type Problem

SUNDAY, OCTOBER 22, 2000
AARMS CATEGORY THEORY SESSION
Location: Chase Building Seminar Room
Chair: Richard Wood

Time	Speaker and Title
9:00am - 10:00am	Jim Lambek, McGill University
	TBA
10:00am - 10:30am	Bob Pare, Dalhousie University
	Orthogonal Adjoints in Double Categories
10:30am - 11:00am	Coffee Break
11:00am - 11:30pm	R. J. Wood, Dalhousie University
	"All" Exactness Conditions Arise from Distributive Laws
11:30pm -12:00pm	Bob Rosebrugh, Mount Allison University
	View Updates for the Sketch Data Model
12:00pm - 12:30pm	Robert Dawson, St. Mary's University
	The Relative Word Problem for Free Extensions of Double
	Categories is Undecidable
12:30pm - 1:00pm	Dorette Pronk, Dalhousie University
	Orbifolds and Quantum Field Theory

SUNDAY, OCTOBER 22, 2000
AARMS BAYESIAN STATISTICS SESSION
Location: Chase Building Colloquium Room
Chair: Chris Field

9:30am - 10:20am	Larry Bretthorst, Washington University
	Nonuniform Sampling, Bandwidth and Aliasing
10:30am - 11:20am	George Gabor, Dalhousie University
	What is the Question?

Conference Abstracts
INVITED LECTURES
Relative Gaps Between Highly Composite Numbers
Morven Gentleman
Dalhousie University

The problem of bounding the relative gap between successive highly composite numbers had its origin in designing Fast Fourier Transform programs to analyze experimental data. Although the observed data sequence could often be of any length which the experimenter chose, the Fourier Transform could only be computed for lengths which were highly composite and were supported by the software, leading the analyst either to throw away data or to have to pad the observed sequence with artificial data (typically zeros) in order to reach one of the supported lengths. The latter approach, though often used, can introduce subtle artifacts into such computed quantities as the estimated power spectrum, and the former approach wastes possibly expensive experimental observations. For conventional Fast Fourier Transform programs that only support lengths that are powers of 2, up to half the data might be discarded. Designing the software to support lengths which are composite numbers with many prime factors could help reduce the fraction of data that must be discarded. To limit the wastage to 10 1,000,000, are factors 2 and 3 enough? Would factors 2, 3, and 5 be significantly better? What about factors 2, 3, 5, and 7?

The naive way to assess the wastage would be to compute all composite numbers of the suggested form within the specified interval, sort them, then look for the largest relative gap between successive numbers of that form. At the time this problem arose, however, this approach was infeasible for the cases of interest because there were so many numbers of the form. The challenge, then, was to compute successive composite numbers of the form with minimal additional working store. This talk will discuss that challenge, for

$$\left \begin{array}{ccccc} & i & j & k & l \\ N=& 2 & 3 & 5 & 7 \end{array} \right.$$

Bayesian Spectrum Analysis and Parameter Estimation
Larry Bretthorst
Washington University

In this tutorial I will try to give people a feeling of what Bayesian probability theory is, and how it differs from the orthodox or frequency interpretation of probability theory. The use of Bayesian probability theory for parameter estimation will be illustrated with a simple example, estimating the frequency and decay rate constant of a decaying sinusoid. In this toy example we derive the posterior probability for the frequency and then show how the sufficient statistic is related to the discrete Fourier transform power spectrum. By solving this simple problem we learn a great deal about both parameter estimation and when the discrete Fourier transform is an optimal frequency estimator, and when it breaks down, why it breaks down.

Impact of Technology on Society
Jacob Slonim
Dalhousie University

The talk will cover issues of the impact of technology on society. In the near future there will specifically be the need for change from technology push to technology pull with special concerns of the have/have not. One of the major impacts of schedule technology, which is changing the way we are going to live in the future, has to do with the concept of electronic commerce. In the next few years we will see a major shift from people physically buying goods and services to people buying them electronically. T his shift has great implications on society and economies and threatens our sovereignty. The shift from a local economy to global economy (also known as the new economy) has great impact on skills and potential changes in both education and lifestyle.

MATH/STAT STUDENT PAPER SESSION
Tutte Connectivity in Graphs
Cedric Davies
Saint Mary's University

A graph is k-connected if when fewer than k vertices are removed the resulting graph is still connected. Tutte uses a slightly stronger definition of connectivity that generalizes to matroids and holds under duality. Inductive tools will be proved for the 2,3-connected cases. An application will be provided by proving an easy corollary to Kuratowski's theorem.

Area and Perimeter Approximation of Circles within a Square
Alasdair Graham
Memorial University of Newfoundland

We state and describe a brief history of the problem, then we describe a program that finds an approximate solution of the problem. Finally, we discuss the problem mathematically.

The Predicted Sequential Regularization Method of
Differential-Algebraic Equations
Colin Macdonald
Acadia University

My talk concerns the investigation of numerical methods for solving differential-algebraic equations (DAEs). In particular, I will examine two new numerical methods called the Sequential Regularization Method (SRM) and the related Predicted Sequential Regularization Method (PSRM).

I will begin by showing that it is easy to construct naive numerical methods for the solution of DAEs; but unfortunately, they often do not work particularly well. Through some simple examples, I will demonstrate and discuss the advantages of the SRM and PSRM. Finally, I will look at the importance of step-size control and how it applies to the SRM and PSRM.

Modelling Electrical Activity in the Human Heart
Mary Maclachlan
Acadia University

Electrical activity in human ventricular muscle can be modelled by the Luo-Rudy gating equations for the most significant ionic currents coupled with a quasi-linear parabolic partial differential equation. The partial differential equation captures the anisotropic conductivity of the muscle fibres: the conductivity is greater along the fibre directions than in the transverse directions. We will show the results of some simulations of this system on an annular region that represents a cross-section through the ventricle.

Cyclically Indecomposable Two-fold Triple Systems from Skolem Sequences
Sarah Mathews
Memorial University of Newfoundland

We describe a research in progress, investigating the construction of cyclically indecomposable cyclic two-fold triple systems from two-fold Skolem sequences. An example of a two-fold Skolem sequence of order 3 is 3, 1, 1, 3, 2, 3, 2, 2, 3, 2, 1, 1. We introduce a modified backtracking algorithm which determines the decomposability of any balanced incomplete block design.

Extended Langford Sequences
Christopher Pollock
Mount Allison University

Extended Langford sequences are generalizations of the more widely known Skolem sequences. While necessary conditions are known for their existence, it is still unknown whether these conditions are also sufficient. I will discuss some new constructions and summarize what is known to date.

A Generalization to the Quantum Binomial Theorem
Ian Rutherford
Mount Allison University

In this talk, I will review the definition of the the quantum binomial coefficients and quantum binomial theorem and prove a generalization.

8-Cycle Decompositions of the Cartesian Product of Two Complete Graphs
Robin Swain
Memorial University of Newfoundland

In this presentation I will be discussing the research on cycle decompositions of graphs that I studied this past summer. In particular, the necessary and sufficient conditions on $m$ and $n$ for $K_m\times K_n,$ the cartesian product of two complete graphs, to be decomposable into cycles of length eight are presented as the main result. A short description of the approach to the proofs for the various cases is outlined, and a complete classification of the leaves which are possible with maximum packings of complete graphs with $8$-cycles is also covered.

APPL MATH & NUMER ANAL SESSION
Multicomponent Gas Transport in Fuel Cell Electrodes
John Stockie
University of New Brunswick

The flow of multicomponent gas mixtures through porous media is encountered in many physical phenomena, ranging from groundwater transport to cell biology. We focus on an application to proton exchange membrane fuel cells, where the "gas diffusion electrode" is a highly porous material that serves to distribute reactant gases uniformly to the active catalyst sites. We develop a mathematical model that is comprised of a coupled system of nonlinear parabolic differential equations. An asymptotic analysis yields a reduced system of equations that captures the slow, diffusively-driven relaxation to steady state at each electrode. The analytical results are compared with computations of the full system, and we also present simulations that demonstrate how fuel cell performance can be optimized by varying electrode geometry and material parameters.

A Delay-Differential Equation Model of Cell-to-Cell Spread of HIV
Rebecca Culshaw
Dalhousie University

We consider a two-dimensional differential equation model of cell-to-cell spread of HIV in tissue cultures, assuming infection is spread directly from infected to healthy cells and neglecting the effects of free virus. Our system is a predator-prey system with infected cells as predators.

We model the intracellular latent period by a time delay. We present stability results for the three equilibria - healthy, trivial and infected - and show how inclusion of a time delay affects stability of the infected equilibrium.

A Stacking Algorithm for Solving Tri-diagonal Symmetric Toeplitz
Systems of Linear Equations
Jeff McNally
University of New Brunswick at Saint John

Systems of linear equations which have symmetric tri-diagonal Toeplitz coefficient matrices have long been studied in many areas of research. In 1999 a work was introduced by the author was introduced that developed a parallel implementation of an inherently sequential problem through the use of a split-correct parallel algorithm. In this work a new approach is presented for solving these systems through the use of a stacking algorithm. In this new algorithm, solutions to specific similar systems are `cropped' and `stacked', thus producing a solution to the original problem. Computer results obtained by implementing the algorithms on a 9 cluster Beowolf using both LAM-MPI and PVM are presented.

Parallel Projection Methods on a Beowulf Cluster
Jing Zhang
University of New Brunswick at Saint John

Most of the existing practical iterative techniques for solving large linear systems of equations utilize a projection process in one way or another. A projection method represents a canonical way for extracting an approximation to the solution of a linear system from a subspace. The coefficient matrix is assumed to be nonsingular tridiagonal and Toeplitz.Such systems appear in many applications with a positive definite or diagonally dominant matrix. Here, through a set of perturbations, we split the original system into several subsystems which can be solved by the parallel projection methods with corrections to follow. The parallel projection algorithms for such special systems will be modified and implemented on a parallel distributed memory computer(a Beowulf cluster).

Solving Banded Linear Systems
Sarita S. Nemani
University of New Brunswick at Saint John

A review of some problems related to solving linear systems whose coefficient matrix is large, sparse, Toeplitz and banded will be considered. In order to do this, first a method for tridiagonal Toeplitz system will be given. Additional remarks will indicate possible extensions to more general problems.

Strong Periodic Forcing in a Predator-prey Model
Andy Foster
Memorial University of Newfoundland

I will discuss the behavior resulting from imposition of seasonal effort harvesting on a species in a continuous predator-prey model (Rosenzweig-MacArthur). Large amplitude periodic harvesting in such population models typically results in richly diverse and complicated bifurcation structures involving chaos. Exhaustive analysis of these structures is impossible at present, but they can be understood through numerical studies and classification of generic bifurcation sequences. Some implications for the prediction and management of real biological populations will be described.

A Simple SIS Epidemic Model with a Backward Bifurcation
James Watmough
University of New Brunswick

In this talk I will formulate an nonlinear Volterra integral equation to model the dynamics of a simple disease. The model always possesses a disease free equilibrium (zero infectives) which loses stability as a parameter (the basic reproduction number) is increased through a threshold. In classical models, this bifurcation is always super-critical. Meaning that the disease always dies out for subthreshold parameter values, and always approaches an endemic level for suprathreshold parameter values. However, with the introduction of a non-constant contact rate, the bifurcation may be sub-critical, implying the existence of multiple equilibrium and hysteresis.

An Embedding Technique for Monotonicity in a Neural Network with Delay
Xingfu Zou
Memorial University of Newfoundland

In a delayed Hopfield neural network that is strongly connected with non-inhibitory interconnections, fast and inhibitory self-connections lead to global convergence to a unique equilibrium of the network. By applying monotone dynamical systems theory and an embedding technique, we prove that this conclusion remains true without the requirement of strong connectivity and non-inhibitory interconnections.

The Numerical Solution of Volterra Integro-differential Equations:
a Survey of Current Research and Open Problems
Hermann Brunner
Memorial University of Newfoundland

Integro-differential of Volterra type and related functional differential equations (e.g. equations with delay arguments) play an important role as mathematical models for biological and physical phenomena with memory effects. In this talk I shall describe some of the current work on the numerical analysis and the efficient computation of solutions of such initial-value problems. The talk will also show that many challenging problems, especially for (partial) VIDEs with nonsmooth solutions or for equations with state-dependent delays, remain to be solved.

CS STUDENT/CONTRIBUTED PAPER SESSION
Toward Faceted Classification of Software Design Models
Fauzi Musbah Ali
University of New Brunswick

Locating and retrieving reusable elements from a large collection of software artifacts in one of the critical problems of software reuse. The most effective way to make the retrieval much easier is to establish a classification scheme for indexing software artifacts. This paper describes and justifies the utilization of faceted classification approach to classifying and retrieving software artifacts at the design level of the description criteria to classify design models for reuse. It also defines the classification model. This work is the first step toward constructing a faceted classification scheme for software design models.

Time-delay Neural Networks designed using Genetic Algorithms
Paul Mountford and Pawan Lingras
Saint Mary's University

Today's data acquisitions systems have the ability to collect vast amounts of data. This data is crucial to a variety of modeling systems. Even though the speed of computer hardware has made remarkable progress in recent years, the current models can no t incorporate all the data that is available from various sources. In fact with some modeling techniques, like neural networks, excessive inputs can confuse the network or reduce its the ability to generalize. It is therefore necessary to reduce the set of input variables. Often an expert will examine the variables and try to reduce their number based on his experience. This can be time consuming at best, and in the worst case, the model's accuracy of prediction can be reduced if important inputs are arbitrarily discarded. The objective of this research was to test the feasibility of using genetic algorithms (GAs) to select a smaller set of appropriate variables to use as input to modeling techniques. Specifically, GAs will be used to determine the connections from the input layer of a time-delay neural network. The techniques developed were applied to two different sets of time series. The first time-series involved monthly asthma hospitalizations. Since the data set was relatively small, the GAs solution was compared the optimal solution obtained using an exhaustive search, as well as previous models developed by other researchers. The second time-series analysis was related to short-term traffic predictions used in intelligent transportation sys tems. The available data was significantly larger than the first data set. Therefore, the results were only compared with the previous models that were formulated based on intuitive understanding of the time-series.

An SQL-based Self-modifying Intranet Search Engine
Colin Cherry and André Trudel
Acadia University

We present an intelligent search engine for a specially indexed collection of documents. It keeps track of all successful queries and the frequency of record and keyword usage. Using this information, the engine improves its accuracy and performance over time.

A Virtual Live Learning Room for First Year
Denis Riordan
Dalhousie University

The Faculty of Computer Science at Dalhousie runs a (real) Learning Center that is well utilized by students. The Center is staffed for more than 12 hours per day with up to six teaching assistants working at peak periods. It services students at all levels on all topics in computer science. The Center consists of 15 stations in a work area. Teaching assistants circulate, helping students on request. This talk describes an evolving attempt to provide a virtual live learning center with analogous properties that is specialized to help first year students with Java programming problems.

In essence the (virtual) Live Learning Room consists of a very user friendly integrated Java development environment centered round a live learning room style communicator. Students in the Live Learning Room can for example, "put up their hands" to ask for help from a professor. Sarah, the room manager, advises whether or not a professor is available. Teachers in the Live Learning Room can "look over a student's shoulder" to view to the work, if need be compiling and running the candidate program, and can give advice either discretely or publicly. Students can chat to others in the Live Learning Room at will. The (virtual) learning room is monitored by a (real) teaching assistant for 16 hours per day and is experiencing increasing usage. The system, which is written in Java, gathers usage information that is being used to improve the learning experience that it provides.

Towards a Logical Basis for Modeling and Querying Distributed Multidimensional Information Systems
Weichang Du
University of New Brunswick

This paper presents a formalism for modeling and querying large-scale distributed information systems with multidimensional structures. It is supported by a rule-based system based on a multidimensional logic ML(Omega). We demonstrate that the rule based system can be used as a deductive front-end to a geographically distributed information system, such as the Web based information systems, and can be used as an aid in information distribution and retrieval of such systems. When coupled with a logic programming language, the system will allow the user specify information distribution and construct queries from different logical and physical perspectives.

An Experiment in Evaluating Automatically Qenerated Hypertext
Jamie Blustein
Dalhousie University

My overall objective is to develop and evaluate ways of automatically incorporating hypertext links into pre-existing scholarly journal articles. I discuss the method I used to generate the links, the evaluation method I developed, and some of the results.

I describe a rule-based approach for making three types of links (structural, definition, and semantic). Structural links are a way of making explicit some connections between parts of the text. Definition links connect the use of a term, defined elsewhere in the document, to that definition. Links that connect parts of text that discuss similar things are semantic links. Several types of semantic links are distinguished.

Two information retrieval (IR) systems (Cornell's SMART system and Bellcore's Latent Semantic Indexing) were employed to select links based on the content of the articles. An experiment was conducted to compare the performance of the links forged using these two systems.

The effectiveness of the links (and the rules used to make them) was tested by people reading the hypertext versions for information under a time constraint. A within-subjects experimental design was used. Each of the nineteen experimental participants read one version of each of three scholarly articles in a different hypertext form (one had only simple links, the others had definition links and semantic links selected using one of the IR systems). Subjects' preferences were also measured.

Corporate IT Architectures
Nur Zincir-Heywood
Dalhousie University

The alliance between computing and communication, as facilitated by the Internet, has lead to a significant change to the way in which computing is perceived, with repercussions to business, industry and service sectors. The traditional model of a computer as a data processing device is giving way to personal digital assistants, web enabled televisions and other devices that rely on communication capabilities. No matter what term is used to describe such phenomenon, the undeniable fact is that information has become a commodity in which networks have an intrinsic role to play. This highlights two issues, firstly, electronic commerce represents a new business opportunity, and hence, secondly, the automatic retrieval of information becomes very important. Indeed, these issues have also become very critical for corporate productivity. In today's world, doing business on the net yields down to being able to define, present, search and retrieve information. As a result, the underlying network should be reliable and accessible for 24/ 7. This means that it is very important to manage and control the corporate information system and network. In the light of these issues, this presentation will emphasize the strategic use of information Technology to capture and deliver knowledge more efficiently and to create a competitive advantage for business.

Problem Solving from Nature
Malcolm Heywood
Dalhousie University

Machine learning has many forms, but at its core often simplifies to conducting some form of search. The purpose of this presentation is to provide a very brief snap-shot of some biologically motivated methods for efficiently conducting such a search. Specific examples include:

Ant Colonies, where this has proved particularly applicable to shortest/ fastest route finding in telecommunication networks;

Evolutionary Computation, a system motivated by Darwin's "survival of the fittest" metaphor and biologically motivated methods for generating new solutions - crossover and mutation. This has received a lot of success in optimization (a search for maximum or minim) and scheduling (solving manufacturing resource association problems).

Neural Networks, takes a very simple neurological model to produce a highly parallel architecture, capable of universal approximation between input and output. This makes the technique particularly applicable to learning "control actions" for problems in which the learning system has to learn "on the job".

There are of coarse may more examples, but we will concentrate on these three, using an application orientated presentation, to provide a taste for what is possible using very simple "natural" models of computation.

MATH/STAT CONTRIBUTED PAPER SESSION
Dynamic Stability of Evolutionary Games
Joe Apaloo
St. Francis Xavier University

Not long after the introduction of the evolutionary stable strategy (ESS) concept, it was noticed that dynamic selection did not always lead to the establishment of the ESS. The concept of continuously stable strategy (CSS) was thereafter developed. It was generally accepted that dynamic selection leads to the establishment of an ESS if it is a CSS. The neighborhood invader strategy (NIS) concept has received very little attention in the literature. Analysis of this evolutionary stability concept shows that it may be impossible for an ESS to be established through dynamic selection even if it is a CSS and no polymorphisms occur. We will examine the NIS concept if it is a CSS and no polymohisms occur. We will examine the NIS concept and its implications for dynamic evolutionary stability of two differential evolutionary games.

New Tilings of the Sphere with Congruent Triangles
Robert Dawson
Saint Mary's University

The edge-to-edge triangulations of the sphere with congruent triangles were classified in 1923 by Sommerville (and again, independently, by Davies in the 1960's.) Recently, I have been considering tilings in which the edge-to-edge condition is relaxed. This yields a number of new tilings, with unusual symmetry groups. I will give several examples of these and discuss briefly the methods used in classifying them.

An Interval Colouring Approach to the Computer Test Scheduling Problem
Megan Dewar
Dalhousie University

Computer chip test scheduling is a growing area of computer science. A new method for developing test schedules, based on interval graph colouring, is introduced. We consider graphs whose vertices are lists and whose vertex adjacency relation is the non-empty intersection of these lists. Some basic colouring results will be presented. The main body of the talk considers weighted random subset graphs. Methods and algorithms for interval colouring these graphs are presented. In particular, an orientation based tabu search is introduced which, when implemented in conjunction with a greedy interval colouring algorithm, yields optimal or near-optimal test schedules.

Polycharacters of Cocommutative Hopf Algebras
M. Kotchetov
Memorial University of Newfoundland

The proposed outline of my talk is the following: 1) Motivation: generalized Lie structures over Hopf algebras, 2)The main theorem: an extension of a well-known M.Scheunert's theorem on skew-symmetric bicharacters of groups to the case of arbitrary cocommutative Hopf algebras of characteristic not 2, 3)The classification of polycharacters on (restricted) enveloping algebras and of bicharacters on divided power algebras.

Introduction to the Algebraic Topology of $C^*-$Extensions
Dan Kucerovsky
University of New Brunswick

Extensions of $C^*-$algebras have turned out to be important in the classification theory for $C^*-$algebras. We explain how $C^*-$extensions are used to form a group by means of so-called "absorbing extensions." If there is time, we will give a theorem characterizing absorbing extensions.

Stabilization and Controllability for
the Transmission Wave Equation
Weijiu Liu
Dalhousie University

In 1980's, Lions proved that if waves pass from a medium domain in which the speed is $a_1$ into a medium domain in which the speed $a_2$ is less than $a_1$ (e.g., from air into glass), then they can be controlled along the exterior boundary of the domain, leaving the case where $a_1< a_2$ as an open problem In this paper we show that they can be still controlled in the case where $a_1< a_2$ by introducing both boundary control along the exterior boundary and distributed control near the transmission boundary and give a physical explanation that it is necessary to introduce the additional control near the transmission boundary.

On a Particular Sum of Dependent Bernoulli and
its Relationship to a Matching Type Problem
Eric Marchand
University of New Brunswick - Fredericton

Let $S_n=\sum_{k=1}^{n} X_k X_{k+1}\;$ and $S=\lim_{n\to\infty}S_n$ where $\{X_k\}_{k=1}^{\infty}$ are independent Bernoulli random variables with mean $p_k$. We study the distributions of $S$ and $S_n$ by establishing a recurrence for the probability generating functions of $S_n$. For the cases when $p_k=\frac{1}{k+B}$ with $B\geq 0$, we show that the distribution of $S$ is a Beta mixture of Poisson distributions. In particular, when $p_k=\frac{1}{k}$, $S$ follows a Poisson distribution with mean 1. We also give an interesting connection with a matching type problem, giving an independent derivation of the above results when $p_k=\frac{1}{k+B}$ with $B$ a nonnegative integer. The talk involves elementary probability and should be accessible to all.

AARMS SESSION-CATEGORY THEORY
Orthogonal Adjoints in Double Categories
Bob Paré
Dalhousie University

We define what it means for a vertical arrow to be adjoint to a horizontal one in a double category. We prove the basic lemmas about them and give examples from ``nature''.

``All'' Exactness Conditions Arise from Distributive Laws
R. J. Wood
Dalhousie University

The title is intended, unprecociously, in the spirit of a famous slogan of the early days of Category Theory: `"All" diagrams commute'. It represents an on-going attempt to refine a framework so as to include many examples in a precise mathematical theorem. Distributive laws, in the sense of Beck, $r:UD\longrightarrow DU$, where $U$ is a coKZ doctrine and $D$ is a KZ doctrine, will be studied with a view to regarding their algebras as categories with certain limits, $U$, and certain colimits, $D$, satisfying an exactness condition, $r$.

View Updates for the Sketch Data Model
Bob Rosebrugh
Mount Allison University

The sketch data model provides a rich environment for semantic data modelling, and provides a clear definition of database views and the View Update Problem. In recent joint work with Michael Johnson, we have made progress on analyzing and resolving view updatability in several important cases. The talk will describe the model and the results obtained.

The Relative Word Problem for Free Extensions of Double Categories is Undecidable
Robert Dawson
St. Mary's University

While the word problem for groups is (in)famously undecidable, the structure of a free extension $G[x]$ follows more or less trivially from that of the original group $G$. Thus, the ``relative word problem'' for free extensions of groups is decidable. If we have an oracle for the word problem on $G$, we can solve the word problem on $G[x]$ in time $O(n+m)$ where $n,m$ are the lengths of the two words.

Something similar happens with many other algebraic structures, including rings (slightly slower: polynomials must be expanded and sorted) and categories.

We shall show that this is not true in double categories; the relative word problem for ${\bf D}[X]$ may be undecidable even when $D$ is finitely presented and has a linear-time algorithm for its word problem.

Orbifolds and Quantum Field Theory
Dorette Pronk
Dalhousie University

I will discuss several different representations of orbifolds and how they can be used to define and calculate orbifold cohomology. I will also give an example of how orbifold cohomology can be used to obtain an example of a quantum field theory.

AARMS SESSION-BAYESIAN STATISTICS
Nonuniform Sampling, Bandwidth and Aliasing
Larry Bretthorst
Washington University

In this talk we will see how Bayesian probability theory leads naturally to a generalized sufficient statistic that in the appropriate limits reduces to the Lomb-Scargle periodogram, the Schuster periodogram, a weighted power spectrum and then finally to a generalized Lomb-Scargle type of periodogram. Additionally, we will see the exact limits of applicability of each of these statistics and we will see what happens to aliases, the parameter estimates, and finally why the generalized discrete Fourier transform that is derived can have multiple peaks in data that are known to contain only a single sinusoid.

What is the Question?
George Gabor
Dalhousie University

Common wisdom in statistics circles holds that orthodox (classical, frequentist, Neyman-Pearsonian) statistics differs from its Bayesian counterpart in that in the former a parameter is a fixed but unknown constant, and as such it is not entitled to a probability distribution, while in the latter it is a random variable equipped with a proper probability distribution and all it entails. It appears that it is this difference in the epistemic status conferred on the data and the parameter that renders the two views irreconcilable: Orthodoxians either reject Bayesian inference outright, or consider its use limited, a sometimes convenient expedient to be used with caution, while Bayesians reject orthodox methods as a collection of ad hoceries, but allow some as convenient expedients to be used with caution, usually only if they have Bayesian interpretation. Is this really at the root of the disagreement? Are these views really irreconcilable? And if not, which view is right? In this talk I will show that the root of the disagreement is much more fundamental, lies much deeper than the mere "ideological", that the two views are indeed irreconcilable, and that the Bayesian view is right by a long shot.

• About this document ...