Graduate Student Combinatorics Seminar, Spring 2005

The graduate student combinatorics seminar is organized by graduate students in the math, computer science, and operations research departments. Students (and perhaps other invited speakers) speak about topics in combinatorics that are of interest to the group as a whole. The talks are to be understandable to everyone present and last less than an hour.

The seminar usually meets Mondays at 1:10pm in the Graduate Student Lounge on the 7th floor of the Hill Center. Everyone is invited to attend.

This semester the seminar is also being sponsored by DIMACS, the Center for Discrete Mathematics and Theoretical Computer Science. DIMACS is providing funding for food and help with publicity, as well as access to DIMACS members and visitors to participate in the seminar.

We are currently looking for speakers (including faculty and visitors) for the seminar. If you would like to give a talk, then please contact Bill Cuckler, whose webpage is here.



Date: Monday, January 24, 2005
Speaker: Paul Ellis
Title: Fodor's Lemma
Abstract: I will state and prove Fodor's lemma, an interesting and useful bit of Set Theory. I will introduce all relavent notions, but familiarity with ordinal numbers would be nice. I will prove one application, and state a surprising application to Graph Theory. I promise that the material is combinatorial in nature.

Date: Monday, January 31, 2005
Speaker: Bruce Sagan, Michigan State University
Title: Proper Partitions of a Polygon and Catalan numbers
Abstract: It is well known that the Catalan numbers count triangulations of a polygon. If one colors the vertices of the polygon and considers triangulations which are proper (in a sense similar to that of proper colorings in Graph Theory) then James Propp conjectured similar enumeration formulas. We prove Propp's conjectures and also discuss how one can move from one proper partition to another by a sequence of moves called flips.

Date: Monday, February 7, 2005
Speaker: Liviu Ilinca
Title: Combinatorial Nullstellensatz
Abstract: I will talk about a version of Hilbert's Nullstellensatz theorem and some of its applications in Combinatorial Number Theory and Graph Theory. This will include the classical Cauchy-Davenport theorem, a proof of a conjecture by Erdos and Heillbron and a result by Alon and Furedi on covering cubes by hyperplanes. I will be following a paper by Noga Alon.

Date: Monday, February 14, 2005
Speaker: Vince Vatter
Title: The structure of permutation ideals
Abstract: Ideals (aka downsets or closed sets) of permutations arise naturally from the analysis of sorting machines and have been extensively studied by enumerative combinatorialists for over two decades now. I will survey the relatively recent research into the structural properties of these sets.

Date: Monday, February 21, 2005
Speaker: Mike Neiman
Title: A partial proof of Ryser's conjecture
Abstract: Ryser's conjecture is a generalization of Konig's theorem to r-partite r-uniform hypergraphs. I will present a proof of the conjecture for the case r=3. This result, due to Aharoni, uses a generalization of Hall's theorem whose proof is topological.

Date: Monday, February 28, 2005
Speaker: Ian Levitt
Title: On a paper of Bourgain
Abstract: In 1990, Jean Bourgain showed that for $A, B \subset [N]$ the sumset $A+B$ contains an arithmetic progression of length $e^{c(log n)^{1/3}}$, where $c$ depends only on the densities of $A$ and $B$. The result itself is significant, the method of proof is even more so. While Bourgain's six pages of mathematical precision are impenetrable to my own powers of deduction, Ben Green is smarter and writes nice exposition papers. I will present the highlights of this proof and some related topics, including a recent result of Chang and subsequent improvement of Bourgain's result by Green.

Date: Monday, March 7, 2005
Speaker: Nick Weininger
Title: Random Linear Extensions of Posets
Abstract: Given a finite partially ordered set P and elements x,y of P, one may choose a linear extension of P (a linear ordering consistent with P) uniformly at random, and evaluate Pr(x > y) in the resultant distribution. The famous "1/3-2/3 conjecture", open for 35 years, says that for any P not a chain, there exist x,y with 1/3 \leq Pr(x > y) \leq 2/3. The first nontrivial result in this direction was proved by Kahn and Saks in 1984; the current best result is due to Brightwell, Felsner, and Trotter from 1995. They show that there exist x,y with (5 - \sqrt{5})/10 \leq Pr(x > y) \leq (5 + \sqrt{5})/10 In a natural and surprising sense, this result is tight for countably infinite posets. Furthermore, the very clever proof of Brightwell et al. leads to a different, simple and far-reaching combinatorial conjecture about linear extensions. We will sketch the proof and state the conjecture.

Date: Monday, March 21, 2005
Speaker: Mike Richter
Title: Linear Algebra Methods in Combinatorics say the darndest things
Abstract: Specifically I will state the nonuniform modular RW Theorem, and talk about the consequences of the nonuniform modular RW Theorem, especially results about the chromatic number of the unit distance graph on R^n and hopefully about the disproval of Borsuk's Conjecture.

Date: Monday, March 28, 2005
Speaker: Thomas Zaslavsky, SUNY Binghamton
Title: Biased Expansion Graphs, Dowling Lattices, Tutte's 3-Decomposition, and Associativity
Abstract: A biased expansion Omega of a graph Delta is a kind of branched covering graph of Delta. A simple kind of biased expansion, built from a group, is called a group expansion. Another kind is a multary quasigroup expansion. (A multary, orn-ary, quasigroup is a set with an n-ary operation that is like a group operation but with n arguments and without an analog of associativity.)

If Delta is 3-connected, then Omega is a group expansion. If Delta is only 2-connected, then Omega is built from group expansions and multary quasigroup expansions in analogy to Tutte's 3-decomposition of graphs.

Biased expansions have associated matroids; those of group expansions of complete graphs are Dowling's lattices. If Delta is 2-connected and not extendible on the same vertex set, then it is built out of Dowling lattices and irreducible multary quasigroup expansions.

The 3-connectedness theorem implies that a multary quasigroup with enough factorizations must be a group in disguise; this proves a conjecture of Belousov. If it does not have enough factorizations, it is obtained by composition of multary operations from disguised groups and irreducible multary quasigroups.

Date: Monday, April 4, 2005
Speaker: Siddhartha Sahi
Title: Higher Correlations and FKG
Abstract: The FKG inequality is a correlation inequality for pairs of random variables defined on a lattice. An interesting consequence of the inequality is that in the random graph G(n, p) monotone increasing events are positively correlated, so knowing that there is a path from x to y makes it more likely that there is a path from v to w. I will give an extension of the FKG inequality for correlations involving more than two random variables. The proof employs formal power series.

Date: Monday, April 11, 2005
Speaker: Sujith Vijay
Title: Variations on a Question of Erdos
Abstract: Given a real number x and positive integers a and b, the quasi-progression Q(x;a,b) is the sequence of integers [ax],[(a+1)x],...,[bx] where [.] denotes the floor function. When x is an integer, the quasi-progression reduces to an ordinary arithmetic progression. Erdos asked in the 1930s whether the non-negative integers can be coloured red and blue such that all arithmetic progressions starting with 0 have bounded discrepancy (i.e., the difference in the number of reds and blues in any arithmetic progression is bounded by some absolute constant). I will use a theorem of Roth to show that if the integers from 0 to N are 2-coloured, there exists a quasi-progression Q(x;a,b) such that x > 2 and Q(x;a,b) has discrepancy at least 0.014 N^(1/6). I will also introduce a game-theoretic "Maker-Breaker" version of the problem, and show how Maker can force a discrepancy of N^(1/2-epsilon) on some arithmetic progression, for sufficiently large N.

Date: Monday, April 18, 2005
Speaker: Drew Sills
Title: Computer proofs via the algorithms of Sister Celine and Zeilberger
Abstract: A series $\sum_{k\geq 0} c_k$ is called hypergeometric if the term ratio $c_{k+1}/c_k$ is a fixed rational function of $k$ for all $k$. Hypergeometric series had been studied extensively throughout the 19th and early 20th centuries, and it turns out that most of the functions of elementary calculus have representations as hypergeometric series. In recent years, there has been a great deal of renewed interest in hypergeometric series and their generalizations. In her 1946 Ph.D. thesis, Sister Mary Celine Fasenmyer provided the first systematic (in fact algorithmic) approach to finding recurrence relations satisfied by hypergeometric summands which allow the automatic derivation of hypergemetric summation formulas. Sister Celine's algorithm, while easy to implement in modern computer algebra systems, is unfortunately rather slow. In 1990, Doron Zeilberger published his creative telescoping algorithm which accomplishes essentially the same goal as Sister Celine's algorithm, but does so much faster. Zeilberger's algorithm was combined with some insights provided by Herbert Wilf to form the "WZ theory", for which the two shared the 1998 AMS Steele prize for seminal contribution to research. During the summer of 2004, Zeilberger, together with his student Mohamud Mohammed, produced a simplfied version of Zeilberger's algorithm. I plan to discuss and demonstrate the algorithms of Sister Celine and Zeilberger.

Date: Monday, April 25, 2005
Speaker: Diane Maclagen
Title:Counting lattice points in polyhedra
Abstract:The problem of counting the number of lattice points in a polytope arises in several areas, including (combinatorial) representation theory. I will give an introduction to some of the techniques.

Links:
Previous semesters: Fall 2004, Spring 2004, Fall 2003 and Spring 2003.
Rutgers Math Department's Seminar Page.
DIMACS/Computer Science Theory of Computing Seminar (previous semesters).
experimental math.
Princeton discrete math seminar.