Mathematics Colloquium, Fall 2004

Talks are on Thursdays from 1-2 pm in NAC 6113 unless otherwise specified

September 9, room NAC 4113: Peter Neumann (Oxford University), Did Galois deserve to be shot?
Abstract: When Evariste Galois was shot in a duel and died in May 1832, he was only twenty years old. Yet he had already published one important paper on his mathematical researches, and had completed an even more important work which, after some vicissitudes, was published posthumously in 1846. His ideas led to a fundamental change in the way mathematicians approached the study of equations. They led directly to group theory and to what later became known as `abstract' or `modern' algebra. In this lecture I will try to paint a word-picture of Galois' turbulent life and his astonishing achievements.

October 7: Ethan Akin (CCNY), About chaos
Abstract: Among the many definitions of chaos is the concept of "sensitive dependence upon initial conditions". We discuss the history and some of the subtle meaning of this idea.

October 14: George Havas (The University of Queensland), How hard is computing with matrices?
Abstract: The calculation of standard forms for matrices plays an important role in many computations. In 1851 Hermite described a standard form for integer matrices which is triangular. Conversion of a matrix representing a set of linear equations to a triangular form provides a way of solving the equations by back-substitution. This form is important in computational linear algebra and number theory. In 1861 HJS Smith described a standard form for integer matrices which is diagonal. This standard form is important in control theory, digital signal processing and in group theory.
Our aim is to develop effective algorithms for finding standard forms of matrices. In principle it is easy, but it can be difficult if the matrix is large.

October 28: Chuck Miller (University of Melbourne), Computability in algebra and geometry
Abstract: This is an overview of the theory and practice of doing computations about algebraic and geometric objects. I will discuss some of the history and meaning of the theoretical limitations on abstract mathematical computation. But I will also indicate why doing computations with groups and geometric objects is still useful even though it is impossible.
My intention is to assume very little specific knowledge on the part of the audience. Hopefully a wide range of mathematicians and computer scientists will be able to follow the discussion.

November 11: Igor Rivin (Temple University), How to measure an egg
Abstract: We use probabilistic methods to get formulas for computing the surface area and other "integral mean curvatures" of an ellipsoid. The formulas involve hypergeometric functions, but we show that there are very simple approximations which become more and more accurate as the dimension of the ellipsoid becomes bigger.

December 2, room NAC 4113: Olympia Hadjiliadis (Columbia University), Change-point detection in the Brownian motion model with two-sided alternatives and its connection to the gambler's ruin problem with relative wealth perception.
Abstract: In this presentation we address two problems; the problem of change-point detection in the Brownian motion model with multiple alternatives and the gambler's ruin problem with relative wealth perception. Although the two problems appear very different, it turns out that there exists a connection that is carefully examined and discussed. In the former problem, the objective is to detect a change in the constant drift by means of a stopping rule when there are multiple possibilities for such a change. Particular attention is drawn to two-sided alternatives, whereby the two-sided CUSUM stopping rule (2-CUSUM) is employed. In particular, it is shown that within two specific classes of 2-CUSUM rules (the harmonic mean rules and the equalizer rules), a particular choice of 2-CUSUM rules enjoys a very strong asymptotic optimality property. In the latter problem, we review the traditional gambler's ruin problem but assume that the gamblers make their decisions based on the relative change of their wealth as opposed to the absolute change. In other words, gamblers are assumed to quit after their wealth makes a significant upward rally or a significant downward fall. In this setting we compute the probabilities of winning (i.e. quitting on the upward fall) or losing (i.e. quitting on the downward fall). Using the above probabilities one can compute the first moment of a broader class of 2-CUSUM stopping rules than the specific class of harmonic mean rules and thus set the first stepping stone on examining the optimality properties of 2-CUSUM stopping rule beyond the traditional setting of equal thresholds in its respective one-sided CUSUM branches.
 


Spring 2004 talks

Fall 2003 talks