The problem presented in this post was discussed in my recent lecture “New types of order types” in the workshop on discrete convexity and geometry in Budapest, a few weeks ago. The lecture described various results and questions including the question: Is there a notion of “order type” based on Tverberg’s theorem. I hope to come back to these questions and results soon. (The lecture also included a poll among the audience about the Erdos-Szekeres conjecture: the outcome was that 12 believe the conjecture and 4 disbelieve.) The present post is based on conversations with Moshe White.
Tverberg’s theorem is among my favorite mathematical theorems; over the years, we devoted eight posts to results and problems around Tverberg’s theorem and mentioned the theorem in a few other places. In 2008 we devoted a post to seven open problems around Tverberg’s theorem. In the meanwhile, two of the problems were settled and major progress was made on a third one. We devoted two posts (I, II) to the background and proof of Tverberg’s theorem.
Here is a link to my 1999 lecture: “An invitation to Tverberg’s theorem” in the 1999 conference “visions in mathematics towards 2000“. Let me also mention Imre Barany’s new proof of Tverberg’s theorem which includes ingredients from Tverberg’s 1965 proof and yet is quite simple and enlightening.
Tverberg’s theorem
Tverberg Theorem (1965): Let Image may be NSFW.
Clik here to view.
be points in Image may be NSFW.
Clik here to view.
, Image may be NSFW.
Clik here to view.
. Then there is a partition Image may be NSFW.
Clik here to view.
of Image may be NSFW.
Clik here to view.
such that
Image may be NSFW.
Clik here to view.
.
The (much easier) case Image may be NSFW.
Clik here to view.
of Tverberg’s theorem is Radon’s theorem.
The maximum number of Tverberg’s partitions
Before asking about the maximum number of Tverberg’s partition let me mention a well-known problem about the minimum:
Sierksma Conjecture: The number of Tverberg’s Image may be NSFW.
Clik here to view.
-partitions of a set of Image may be NSFW.
Clik here to view.
points in Image may be NSFW.
Clik here to view. is at least Image may be NSFW.
Clik here to view.
. Let’s denote this minimum by Image may be NSFW.
Clik here to view.
.
(Here we consider unordered partitions. For ordered partitions we need to multiply by Image may be NSFW.
Clik here to view.
.) It is quite interesting that there are quite a few different configurations for which this minimum is attained. We discussed many examples by Moshe White and by Boris Bukh, Po-Shen Loh, and Gabriel Nivasch in this post.
What about configurations with the maximum number of Tverberg’s partitions? Here we need to assume that the points are “generic”;
Problem 1: What is the maximum number of Tverberg partitions for (generic) Image may be NSFW.
Clik here to view.
points in Image may be NSFW.
Clik here to view.
. Let us denote the answer by Image may be NSFW.
Clik here to view.
.
Here is a variant of the problem:
Problem 2: Given Image may be NSFW.
Clik here to view.
points in Image may be NSFW.
Clik here to view. in a generic position, what is the maximum number of partitions to Image may be NSFW.
Clik here to view. parts such that the positive hulls will have a point in common. Let us denote the answer by Image may be NSFW.
Clik here to view.
.
(It is reasonable to think that Image may be NSFW.
Clik here to view.
but I don’t have proof.)
The Gaussian model and a lower bound (for a variation of the problem)
Let Image may be NSFW.
Clik here to view. be points in Image may be NSFW.
Clik here to view., Image may be NSFW.
Clik here to view.
, and consider an assignment of signs Image may be NSFW.
Clik here to view.
, Image may be NSFW.
Clik here to view.
. Consider a partition into Image may be NSFW.
Clik here to view. parts Image may be NSFW.
Clik here to view.
such that
Image may be NSFW.
Clik here to view.
We will consider only partitions for which Image may be NSFW.
Clik here to view.
, and let Image may be NSFW.
Clik here to view.
denote the number of such (unordered) partitions. Image may be NSFW.
Clik here to view.
.
Now, let’s choose Image may be NSFW.
Clik here to view.
to be independent Gaussian vectors.
Claim 1: (almost surely) for every such partition Image may be NSFW.
Clik here to view. there are unique signs such that our equalities hold.
Claim 2: The probability for an equality to hold is the same for every choice of signs.
Therefore the expected number of partitions for the all one signs is at least
Image may be NSFW.
Clik here to view.
Clik here to view.
When Image may be NSFW.
Clik here to view.
Clik here to view.
Clik here to view.
Clik here to view.
Image may be NSFW.
Clik here to view.
compared to
Image may be NSFW.
Clik here to view.
that Sierksma’s conjecture gives. It is not outrageous to conjecture that Image may be NSFW.
Clik here to view.
Clik here to view.
Clik here to view.
In the third paper of his Ph. D. thesis, Moshe White had theoretical and empirical results regarding the Radon behavior of Gaussian points in the plane and higher dimensions. (This study inspired the above argument regarding Problem 2.) Let me describe some empirical results regarding the possible behaviors of Tverberg’s partitions for seven points in the plane.
Seven points in the plane
The following picture (by Moshe White) presents the 11 types of configurations of 7 points in the plane, the first non-trivial case of Tverberg’s theorem for three parts. The blue lines between the types correspond to “moves” of Barany’s new proof. There are three configurations with 4 partitions (the minimum) and one configuration with seven partitions (the maximum).
Image may be NSFW.
Clik here to view.
Image may be NSFW.
Clik here to view.
Image may be NSFW.
Clik here to view.