Dept Colloquium Schedule for 2006/2007

Maintained by: Qingtang Jiang

Select:

Fall
Spring
Attendence form (graduate CS students admitted Fall 2005 and later)
Previous colloquium talks in 2005/2006 maintained by: Cezary Z. Janikow

Note: Standard venue is 302CCB, standard day/time is 3:30pm, Thursday.
Tools: Please specify in advance: e.g., overhead or PC projector, viewer, network, etc
Audience: General (faculty and students) or specific (as defined)

Schedule for Fall

Date	Place	Refreshments Time	Talk time	Speaker	Title
9/14 Thursday	302CCB	3:00pm	3:30pm	Kamlesh Parwani	Actions of SL(n,Z) on homology spheres
9/21 Thursday	302CCB	3:00pm	3:30pm	William Connett	The Chebyshev polynomials
9/27 Wednesday	302CCB	3:00pm	3:20pm	Henry Kang	A unified scheme for adaptive stroke-based rendering
10/5 Thursday	302CCB	3:15pm	3:30pm	Ronald Dotzel	A Schur lemma for homotopy representations
10/12 Thursday	302CCB	3:00pm	3:30pm	William Connett	The controversy over the two dimensional analog of the Chebyshev polynomials
10/19 Thursday	302CCB	3:30pm	4:00pm	Charles Doran	String theory and mathematics
10/26 Thursday	302CCB	3:00pm	3:30pm	Akira Sano	Grassmannian and the space of lattices over Witt vectors
11/1 Wednesday	302CCB	3:00pm	3:30pm	Cezary Janikow	Heuristics in genetic programming
11/9 Thursday	302CCB	3:00pm	3:30pm	G.V. Ravindra	Matrices with polynomial entries and equations defining subvarities
11/13 Monday	302CCB	3:30pm	4:00pm	En-Bing Lin	Wavelet-based computational methods and applications in Symplectic Geometry
11/15 Wednesday	302CCB	3:00pm	3:30pm	Pier Luca Lanzi	Learning classifier systems: introduction, recent developments, and future directions

Schedule for Spring

Date Place Refreshments
Time Talk time Speaker Title

3/7
Wednesday
302CCB
2:00pm
2:30pm
Ronald P. Loui
Streaming AI and hardware-based document classification

3/9
Friday
302CCB
1:00pm
1:30pm
Maria Charina
Tight wavelet frames for subdivision

3/12
Monday
302CCB
9:30am
10:00am
Fikret Ercal
Bioinformatics applications in parallel EST clustering and gene family identification

3/12
Monday
302CCB
1:30pm
2:00pm
Marco Peloso
Boundary behaviour of biholomorphic mappings and analysis on the worm domain

3/19
Monday
302CCB
2:00pm
2:30pm
Elizabeth A. Kellogg
Computational problems in DNA sequence analysis

4/4
Wednesday
302CCB
2:00pm
2:30pm
Daniel R. Tauritz
Real-world adversarial game-theoretic problem solving employing competitive coevolutionary armsraces: Critical infrastructure protection & automated software engineering

4/12
Thursday
302CCB
2:00pm
2:30pm
Ronald Dotzel
On the Goldbach conjecture

4/25
Wednesday
302CCB
2:00pm
2:30pm
Mark Hauschild
Analyzing probabilistic models in hierarchical BOA on traps and spin glasses

Speaker: Maria Charina
Title: Tight wavelet frames for subdivision
Abstract: We show how to construct multivariate tight wavelet frame decom- positions for scalar and vector subdivision schemes with nonnegative masks. The constructed frame generators have one vanishing moment and are obtained by factorizing certain positive semi-definite matri- ces. The construction is local and allows us to obtain framelets even in the vicinity of irregular vertices. Constructing tight frames, instead of wavelet bases, we avoid extra computations of the dual masks. In addition, the frame decomposition algorithm is stable as the discrete frame transform is an isometry on l2, if the data are properly normalized. This is joint work with Joachim Stoeckler.
Bio: Dr. Charina received her Ph.D. in Math from Auburn University in 2002. She is currently with the Department of Math at University of Dortmund, Germany. Dr. Charina's reserach interests include the non-tensor-product based constructions of nonstationary tight spline frames on bounded domains, the constructions of divergence-free frames on domains with complicated geometries, convergence and regularity of multivariate vector subdivision schemes, and magnetohydrodynamics and multiscale methods for PDEs.
Audience: General
Tools: PC projector

Speaker: William Connett
Title: The Chebyshev polynomials
Abstract: Polynomials are such useful mathematical objects, that they are almost pervasive in Mathematics. They provide important examples and tools in fields as disparate as algebraic number theory and numerical analysis. I will talk about some interesting individual polynomials and some important families of polynomials. This will allow me to introduce the audience to some of the history and culture of these important mathematical objects. My main interest is in the sociology of polynomials, and in particular families of orthogonal polynomials. The Chebyshev polynomials were historically the first example of such a family, and they are still one of the most important cases.

Title: The controversy over the two dimensional analog of the Chebyshev polynomials
Abstract: There have been a number of efforts to generalize the theory of polynomials from one dimension to n-dimensions. In some cases this has been straight forward, but in other cases this has been surprisingly difficult. First, it is not even clear what it should mean for a family in two real variables to be orthogonal. Second, there are few examples of such families that are not merely the product (or tensor) of two one dimensional examples. Third, the generalization of a classical example, like the Chebyshev polynomials, really depends on what the user wants to do with them. Thus the generalization produced by an analyst might be quite different than the generalization produced by someone who works in representation theory. It is perhaps not surprising that none of these competing generalizations seem to be motivated by the problems that motivated Chebyshev in his discovery of the one dimensional family that bears his name.

Bio: Although William Connett was born in Mexico City in 1939, he grew up in St. Louis. He attended John Burroughs School, Georgetown University, and the University of Chicago where he received his Ph. D. in 1969. His research has been in classical analysis (special functions, singular integrals, multiplier theories, and measure algebras) and numerical analysis (grid generation, flow visualization, and computational fluid dynamics). He is also an enthusiastic statistician, specializing in the applications of statistics to biology. Dr. Connett spent his entire academic career at the University of Missouri in St. Louis, and since 2002 has been a Research Professor (Emeritus). He is the only member of the faculty to have been employed as a caddy at the former Bellerieve Country Club, where the campus is currently located.
Audience: General
Tools: Overhead projector

Speaker: Charles Doran
Title: String theory and mathematics
Abstract: Why are mathematicians so excited about string theory? It is clear why a physicist might be interested in a theory that seeks to unify all the forces of nature, a theory whose stated goal is to obtain a consistent set of equations explaining galaxies and subatomic particles and everything in between. It's not so clear why mathematicians who study geometry, topology, algebra, combinatorics, number theory, etc. should care. The answer lies in the powerful application of ideas from string theory within mathematics. Indeed string theory has pointed out deep connections between whole branches of mathematics that were previously thought unrelated. It is the goal of the speaker to convey the spirit of these developments without the technical details.
Bio: Charles Doran is an Assistant Professor at the University of Washington in Seattle. He received his PhD from Harvard University in 1999 with a thesis written under the direction of Shing-Tung Yau and Barry Mazur. His reseach intersets cover algebraic geometry, number theory and string theory.
Audience: General
Tools: PC projector

Speaker: Ronald Dotzel
Title: A Schur lemma for homotopy representations
Abstract: In this talk, I will discuss some interesting aspects of the representation theory of finite groups and the rather surprising connections with the topology of finite group actions on spheres. A new result in this direction is:
If X is a homotopy representation of a p-group P we say that P acts irreducibly if the dimension function of X corresponds to an irreducible real representation of P. Let C be a cyclic group of prime order q. We then have
Theorem Suppose that X is a homotopy representation of the group PxC and that P acts irreducibly on X. Then the fixed point set of C is either empty or all of X.
The above jargon notwithstanding, this will be a “low-key” talk accessible to our typical colloquium audiences. Really.
Bio: Ronald Dotzel received his Ph. D. from Rutgers University and has been at UMSL since 1980.
Audience: General
Tools:

Speaker: Ronald Dotzel
Title: On the Goldbach conjecture
Abstract: The Goldbach Conjecture originated from a letter to Euler in 1742, written by Christian Goldbach, asking whether Euler could prove that every even number larger than 4 is the sum of two odd primes. Euler replied that he did not know of a proof and that the problem was likely very hard. To this day a proof has not been found. It has been verified for numbers as large as 4*10^11 and is generally believed to be true. I will present an elementary proof that it is true (mod n) for any n larger or equal 2. Namely,
If k is even and n is greater than 1 then there exist odd primes p and q such that p+q k (mod n).
This talk would be of interest to undergraduates as well as others.
Bio: Ronald Dotzel received his Ph. D. from Rutgers University and has been at UMSL since 1980.
Audience: General
Tools:

Speaker: Fikret Ercal
Title: Bioinformatics applications in parallel EST clustering and gene family identification
Abstract: Recent technological advances have greatly increased the production of ESTs (Expressed Sequence Tags) which in turn have facilitated gene discovery in a variety of different species. Although using ESTs is a cost-effective approach in gene discovery, the amount of data and hence the computational resources required make it a very challenging problem. Time and storage requirements for EST clustering and gene family identification problems are prohibitively expensive. Our work involves developing an intelligent, time- and memory-efficient parallel clustering algorithm for the soybean (Glycine max) EST database as well as software tools and novel methods for identifying gene family members. We have developed software applications (SimESTs, PCAT, and SCAT) which simplified the gene family identification by automating its steps and eliminating the need for manual intervention and hence significantly reducing the time taken for the process.
Bio: FIKRET ERCAL received the B.S. (with first-class honors) and M.S. degrees in EE from Istanbul Technical University, and the Ph.D. degree in Computer Science from the Ohio State University in 1988. He is currently a professor and interim chair of the Computer Science Department at UMR. Dr. Ercal is IEEE Golden Core Member and the recipient of Meritorious Service Award from IEEE Computer Society. He has received three Faculty Excellence Awards from UMR and a Best Paper Award. He has served on the editorial boards of several journals, on the UM Research Board, and as the co-Program and co-General chair for the annual BioSP3 workshop. He has numerous publications in the form of books, book chapters, journal and conference articles in the areas of bioinformatics, parallel and distributed computing, and computer vision.
Audience: General
Tools: PC projector

Speaker: Mark Hauschild
Title: Analyzing probabilistic models in hierarchical BOA on traps and spin glasses
Abstract: The hierarchical Bayesian optimization algorithm (hBOA) can solve nearly decomposable and hierarchical problems of bounded difficulty in a robust and scalable manner by building and sampling probabilistic models of promising solutions. This paper analyzes probabilistic models in hBOA on two common test problems: concatenated traps and 2D Ising spin glasses with periodic boundary conditions. We argue that although Bayesian networks with local structures can encode complex probability distributions, analyzing these models in hBOA is relatively straightforward and the results of such analyses may provide practitioners with useful information about their problems. The results show that the probabilistic models in hBOA closely correspond to the structure of the underlying optimization problem, the models do not change significantly in subsequent iterations of BOA, and creating adequate probabilistic models by hand is not straightforward even with complete knowledge of the optimization problem.
Bio: Mark Hauschild is a Teaching Assistant at UMSL and does research for the Missouri Estimation of Distribution Algorithms Laboratory (MEDAL)
Audience: General
Tools: PC projector

Speaker: Cezary Janikow
Title: Heuristics in genetic programming
Abstract: Genetic Programming simulates nature for problem solving, while using trees for its simulated DNAs. The user defines the representation space for a problem by defining the set of functions and terminals which label the nodes in the trees. The sufficiency principle requires that the set be sufficient to build the desired solution trees. To satisfy this principle, the user is often forced to provide a large set, which unfortunately also enlarges the representation space and thus, the search space. Structure-preserving crossover, STGP, CGP, and CFG-based GP, are example methodologies that allow pruning the space by specifying rules for valid or preferable tree construction. CGP is based on the idea of using strong and weak local heuristics to restrict or focus the desired representation space, in an efficient way. In the talk, after introducing GP and its search space, we will discuss various strategies for using heuristics, from global probabilistic tree models to local heuristics used in CGP. We will illustrate how local heuristics improve search properties (time and space), how such heuristics can be discovered, and finally how they can be extended to non-local context.
Bio: Cezary Z. Janikow received B.A. (1986) and M.S. (1987) from the University of North Carolina at Charlotte, and Ph.D. (1991) from the University of North Carolina at Chapel Hill, all in Computer Science. His research interests include evolutionary computation, representation and constraints in genetic algorithms and genetic programming, machine learning, more recently fuzzy representation for machine learning. His other interests include software design, development, and management (both structured and object-oriented), and C/C++ programming. He has been involved in a number of grant- and contract-based projects, reviewed for more than twenty different journals, published about fifty journal and conference papers, and was a proposal reviewer for a number of federal and local agencies.
Audience: General
Tools: PC with PowerPoint, PC projector

Speaker: Henry Kang
Title: A unified scheme for adaptive stroke-based rendering
Abstract: In this talk we present a comprehensive scheme for automatically generating a broad class of artistic illustrations from photographs. Using strokes as the major building blocks, our system optimizes the stroke attributes subject to the desired rendering style. The stroke attributes are computed adaptively to enable importance-based control of the abstraction level at each pixel. We propose a novel outline detection and refinement paradigm to construct an outline map, and from which to derive the pixel-wise importance. We also introduce an adaptive bilateral filter to adaptively guide the curved stroke directions based on the importance map. Given the outline, importance, and direction maps, the system creates the illustration via selecting the representative colors, setting the style parameters, and optimizing the stroke attributes based on simulated annealing. The experimental results show that our scheme facilitates automatic production of artistic illustrations in a wide range of rendering styles.
Bio: Henry Kang is an Assistant Professor of Computer Science at the University of Missouri, St. Louis. He earned his Ph.D. in Computer Science from KAIST (Korea Advanced Institute of Science and Technology) in 2002. His main research interests include non-photorealistic rendering and animation, image-based modeling and rendering, image and video processing, facial expression animation, curve and surface modeling, etc.
Audience: General
Tools: Projector

Speaker: Elizabeth A. Kellogg
Title: Computational problems in DNA sequence analysis
Abstract: My lab is involved in two major areas of bioinformatics. First, we study the evolution of organisms and their genes by comparative analyses of DNA sequences. The general procedure involves assembling sequences of the DNA for a comparable gene (e.g., the gene encoding a particular enzyme) from multiple organisms. Some of these sequences are downloaded from GenBank (the major biological database) and others we generate ourselves. The sequences are then aligned to permit detection of mutations. Finally the aligned genes are submitted to one of several programs for construction of evolutionary trees. Most of these use an optimality criterion (maximum likelihood, Bayesian posterior probability) to determine the "best" tree; the computational problem is NP-complete and thus becomes more demanding as the number of DNA sequences increases. The second bioinformatic problem we address is manipulation of biological databases. As more and more DNA sequences are deposited in GenBank we are able to move away from wet-bench work and into data mining. However, available tools for data mining often are inadequate to automate our most routine tasks. We have developed one such tool, but would welcome input on others. I will outline the general problems we face without describing any particular solutions.
Bio: Elizabeth (Toby) Kellogg is the E. Desmond Lee and Family Professor of Botanical Studies in the Department of Biology at UMSL. She has an associate appointment at the Missouri Botanical Garden, and her lab provides bench space for graduate students from the Garden. She has been at UMSL since 1998, before which she was an Associate Professor at Harvard University.
Audience: General
Tools: LCD projector

Speaker: Pier Luca Lanzi
Title: Learning classifier systems: introduction, recent developments, and future directions
Abstract: Learning classifier systems are methods of machine learning that exploit temporal difference learning and genetic algorithms to solve problems through trial-error interactions. They maintain a population of condition-action-prediction rules (the classifiers) which represents the current solution to the given target problem. Temporal difference learning is used to estimate the goodness of the rules that classifiers in the population. Genetic algorithms are used to search for better classifiers through the selection, recombination, and mutation of existing ones.
In this talk I will give a short introduction to learning classifier systems, overview the recent developments in this area, and discuss the most promising research directions.
Bio: Pier Luca Lanzi is associate professor in Computer Engineering at the Politecnico di Milano, Dept. of Electronics and Information. His research areas include evolutionary computation, reinforcement learning, and machine learning. He is interested in applications to data mining and autonomous agents.
Audience: General
Tools: PC projector

Speaker: En-Bing Lin
Title: Wavelet-based computational methods and applications in Symplectic Geometry
Abstract: We present a general approach to obtain numerical solutions for linear and nonlinear partial differential equations by using wavelet-based methods such as scaling function interpolations and the method of connection coefficients. We also consider applications of wavelet-based methods to nonlinear dynamical problems. In our approach, we take into account of the underlying algebraic, geometric and topological structures of the corresponding problems. More precisely, we present wavelet approach to solve Hamilton problems in Symplectic Geometry.
Bio: En-Bing Lin received Ph. D. (1982) from Johns Hopkins University. He is a Professor at the University of Toledo. He visited M. I. T., UW-Milwaukee and UC-Riverside before joining UT in 1986. His research interests include wavelet analysis and applications, mathematical physics and differential geometry.
Audience: General
Tools: PC with PowerPoint, PC projector

Speaker: Ronald P. Loui
Title: Streaming AI and hardware-based document classification
Abstract: In this talk, we review the origins of our 2000-2002 idea to put classic UNIX text streaming tools on FPGA's. We discuss the originally envisioned applications and the academic significance of the paradigm of "Streaming AI". In this conception of AI, computational resources are allocated to bandwidth maximization, and algorithms that are simple are preferred to those that are complex. Most of the technical discussion will be drawn from an intelligence project for classifying documents in the communications network at line speeds, 2001-2006. Some datamining and information retrieval issues will appear here. We conclude with some speculation as to future applications and some policy ideas.
Bio: Professor Ronald P. Loui has been teaching artificial intelligence and programming methodologies at Washington University in St. Louis for twenty years. His interests have been either very foundational (reinventing the mathematical models of the social sciences to include information processes in games, decisions, risk, and rules), or very practical (co-patenting a content-based firewall, advocating the use of scripting languages and web technologies for rapid development, and warning about the strategic implications of a generation of programmers raised on JAVA).
Audience: General
Tools: PC, PC projector

Speaker: Kamlesh Parwani
Title: Actions of SL(n,Z) on homology spheres
Abstract: Any continuous action of SL(n, Z), where n > 2, on a r-dimensional mod 2 homology sphere factors through a finite group action if r < n? In particular, any continuous action of S L(n + 2, Z) on the n-dimensional sphere factors through a finite group action.
Bio: Kamlesh Parwani is a Visiting Assistant Professor at the University of Missouri--St. Louis. He recieved his PhD at Northwestern University in 2003. His primary interests include dynamical Systems, group actions on manifolds, and low-dimesional topology.
Audience: General
Tools: Projector

Speaker: Marco Peloso
Title: Boundary behaviour of biholomorphic mappings and analysis on the worm domain
Abstract: A question of fundamental importance in analysis and geometry of several complex variables is whether every biholomorphic map between two smooth bouneded domains extends to a C^\infty diffeomorphism of the boundaries. This question is still open. In recent years the so-called worm domain of Diederich and Fornaess has been shown to present many pathologies and to be a potential candidate as a counterexample. In this colloquium talk, I will discuss this problem and some recent progress obtained in collaboration with Steven Krantz, in particular on the analysis of the Bergman kernel of the worm domain, the analysis of its singularities and of its mapping properties.
Bio: Marco Peloso received his Ph.D. degree at Washington University in St. Louis, in 1990. Since 1991 he has been member of the faculty of the Department of Mathematics at the Politecnico di Torino, Italy. His research interests include function theory in several complex variables, harmonic analysis, and partial differential equations.
Audience: General
Tools: PC, PC projector

Speaker: G.V. Ravindra
Title: Matrices with polynomial entries and equations defining subvarities
Abstract: We shall discuss the following two questions.
1. Given a polynomial, when can we say that it (or some power of it) occurs as a determinant of suitable matrix.
2. Given a submanifold Y in a manifold X, under what conditions is Y given by the "correct" number of equations?
Bio: Dr. Ravindra received his Ph.D. from the Institute of Mathematical Sciences, India in 2001. Currently, he is a Chauvenet Lecturer at Washington University in St. Louis.
Audience: General
Tools:

Speaker: Akira Sano
Title: Grassmannian and the space of lattices over Witt vectors
Abstract: The aim of the talk is to discuss some aspects of a geometric object introduced by W. Haboush, an analogue of an affine Grassmannian associated naturally to a certain linear algebraic group G. I will begin with the space of lines as a special case of a finite-dimensional Grassmannian, and move on to its infinite-dimensional analogue in two cases; one of which is Haboush's construction as a space of lattices over a certain complete local field. I will discuss some motivations for (as well as possible applications of) this idea.
Bio: Akira Sano is a Visiting Assistant Professor at the University of Missouri at St. Louis. He received his PhD at University of Illinois at Urbana-Champaign in 2004. His primary research interests include algebraic groups, their homogeneous spaces and their certain generalizations.
Audience: General
Tools: Overhead projector

Speaker: Daniel R. Tauritz
Title: Real-world adversarial game-theoretic problem solving employing competitive coevolutionary armsraces: Critical infrastructure protection & automated software engineering
Abstract: There are many important real-world problems which are of a game-theoretic nature. In game-theoretic problems multiple agents try to optimize their own utility functions which are dependent on other agents. When the agents' utility functions are in opposition to each other - that is, if increasing one's own utility value causes a corresponding decrease in another agent's utility value - then optimizing the utility functions is called an adversarial search problem. Unfortunately, the search spaces associated with real-world adversarial problems are typically astronomically large and ill-behaved, making exhaustive search - even with advanced game tree pruning techniques - utterly infeasible, and also precluding the use of well known local search algorithms such as simulated annealing. However, by casting adversarial search as an a priori armsrace one can employ competitive coevolution to efficiently search for high-quality solutions. This presentation will provide an introduction to competitive coevolutionary algorithms and illustrate their use for real-world problem solving on one problem in critical infrastructure protection, namely infrastructure hardening, and one problem in automated software engineering, namely automated software correction.
Bio: Daniel Tauritz received his M.S. (1996) and Ph.D. (2002) in Computer Science from Leiden University in the Netherlands. He is currently an Assistant Professor of Computer Science at the University of Missouri-Rolla where he teaches a two-course sequence on Evolutionary Algorithms (EAs). Furthermore, he is the Director of the Natural Computation Laboratory, Research Investigator in UMR's Intelligent Systems Center, and campus coordinator for Sandia National Laboratories Center for Cyber Defenders. His research interests are primarily in the area of EAs and their applications for real-world problem solving, currently with an emphasis on developing novel self-adaptive EAs and applying competitive coevolutionary EAs to problems in the domains of critical infrastructure protection and automated software engineering.
Audience: General
Tools: Laptop, PC projector

Speaker:
Title:
Abstract:
Bio:
Audience:
Tools:

Back to the home page of the Department of Mathematics and Computer Science