Raum MA 313 Mathegebäde, TU Berlin Straße des 17. Juni, 136 10623 Berlinmit Ausnahme der Vorlesung von Jesus De Loera am Montag, welche in Raum MA 041 gehalten wird.
Montag, 6. Dezember 2004 |
||
| 9:00 | Colin McDiarmid (U Oxford) | Randomness and the Plane (Part I) |
| 10:30 | Kaffeepause | |
| 11:00 | Mihyun Kang (HU Berlin) | Random Planar Structures |
| 11:45 | Manuel Bodirsky (HU Berlin) | The Asymptotic Number of Cubic Labeled Planar Graphs |
| 12:30 | Mittagessen | |
| 14:15 | Jesus De Loera (UC Davis, U Magdeburg) | The Many Aspects of Counting Lattice Points in Polytopes, (Raum MA 041) |
| (16:00 | Sarah Renkl (TU Berlin) | Colloquium of the Graduate College CGC, Raum MA 041) |
Dienstag, 7. Dezember 2004 |
||
| 9:00 | Colin McDiarmid (U Oxford) | Randomness and the Plane (Part II) |
| 10:30 | Kaffeepause | |
| 11:00 | Arie Koster (ZI Berlin) | Degree-based Treewidth Lower Bounds - Theory and Computation |
| 11:45 | Benjamin Hiller (ZI Berlin) | Probabilistic Competitive Analysis of an Elementary Dial-a-Ride Problem |
| 12:30 | Mittagessen | |
| 14:30 | Jesus De Loera (UC Davis, U Magdeburg) | Generating Function Algorithms for Lattice Point Problems (Part I) |
| 16:00 | Kaffeepause | |
| 16:30 | Raymond Hemmecke (U Magdeburg) | Test sets in integer programming: recent developments |
| 18:30 | gemeinsames Abendessen | |
Mittwoch, 8. Dezember 2004 |
||
| 9:00 | Colin McDiarmid (U Oxford) | Randomness and the Plane (Part III) |
| 10:30 | Kaffeepause | |
| 11:00 | Rafael Mechtel (TU Berlin) | Random Edge on Products of Polygons |
| 11:45 | Frank H. Lutz (TU Berlin) | Enumeration and Random Realization of Triangulated Surfaces |
| 12:30 | Mittagessen | |
| 14:00 | Jesus De Loera (UC Davis, U Magdeburg) | Generating Function Algorithms for Lattice Point Problems (Part II) |
In contrast to general planar graphs, the generating function for the number of connected cubic planar graphs is algebraic. We present a decomposition of cubic planar graphs, derive equations for the generating functions, and analyse their singularities. This allows us to compute the asymptotic number of cubic labeled planar graphs. The talk builds on the previous talk "Random planar structures" by Mihyun Kang.
In this talk we give an introduction to test set methods in integer programming. Although these sets are exponential in size, they can be encoded efficiently via Barvinok's short rational generating functions. Besides theory and algorithms, we present applications of test sets to stochastic integer programming, to nonlinear convex integer programming, to counting lattice points, and to stochastics.
We consider an online variant of a special Dial-a-Ride problem where the transportation network is a tree. Requests are generated by chosing source and destination vertex according to an arbitrary probability distribution on the vertices. We show that an algorithm called IGNORE(use-MST) is aas (1+o(1))-competitive wrt the total distance traveled by the server if there is high load.
We consider random planar structures, such as planar graphs, series-parallel graphs, and outerplanar graphs. We ask the following questions: how many of them are there, can we sample a random instance uniformly at random, and what properties does a random planar structure have.
Our approach to answer these questions is to decompose the graphs along their connectivity. For the exact enumeration and the uniform generation we use the so-called recursive method: We derive recursive counting formulas along the decomposition, which yields a polynomial time algorithm to sample a graph that is uniformly distributed. For the asymptotic enumeration we interpret the decomposition in terms of generating funtions and derive the asymptotic numbers of the graphs, using singularity analysis.
On one hand, having the uniform sampling algorithm, we can discuss empirical properties of random planar structures. On the other hand, having the asymptotic number, we can discuss the asymptotic properties of random planar structures that hold when the number of vertices converges to infnity.
Every lower bound for treewidth can be extended by taking the maximum of the lower bound over all subgraphs or minors. This extension is shown to be a very vital idea for improving treewidth lower bounds. The studied lower bounds have in common that they are all the vertex degree of some vertex in a subgraph or minor of the input graph. We study the computational complexity of these lower bounds and present computational experiments on algorithms that compute the bounds exactly/approximately.
It is, in general, a difficult problem to decide realizability for triangulated orientable surfaces of genus \geq 1$. Surprisingly, for triangulations with few vertices, 3-dimensional geometric realizations (with straight edges, flat triangles, and without self-intersections) can be obtained by choosing coordinates randomly: At least 827 of the 865 vertex-minimal 10-vertex triangulations of the orientable surface of genus 2 are realizable. We present some of these examples in the talk. Moreover, we explain the backtracking approach that we used to enumerate the 42.426 triangulated surfaces with 10 vertices.
The random edge rule used as a pivot rule in the simplex algorithm looks quite
simple: "Given a polytope and a linear objective function, then move from a
given vertex uniformly at random to one of its improving neighbors, until the
optimal vertex is reached."
But despite its simplicity the random edge rule seems to be awful to analyze in
any nontrivial example. This is allready true for such an easy example as the
product of two polygons, which yields a four-dimensional polytope with the number
of vertices being quadratic in the number of facets.
We exploit the combinatorial structure of these polytopes and of
unique-sink-orientations to show that in expectation the random edge rule visits
not more than a linear number of vertices.
The used methods can be extended to analyse four-dimensional dual-cyclic-polytopes.
But they become technically more challenging.
Postanschrift: DFG Research Center MATHEON TU Berlin MA 6-2 Straße des 17. Juni 136 10623 Berlin Germany
Tel. : (+49 30) 314 24604 Fax : (+49 30) 314 21269 e-mail : kaibel@math.tu-berlin.de
Postanschrift Johann von Neumann-Haus Humboldt-Universität zu Berlin Institut für Informatik Rudower Chaussee 25 12489 Berlin
Tel. : (+49 30) 2093 3830 Fax : (+49 30) 2093 3191 e-mail : kang@informatik.hu-berlin.de