AG Diskrete Mathematik und Mathematische Informatik
Winter Term 2007 / 2008

Prof. Dr. C.-P. Schnorr, Prof. Dr. T. Theobald
Dr. H. Bosse, R. Hartung, C. Riener, A. Scemama, R. Steffens



Semidefinite programming bounds for spherical codes

Lecture by F. Vallentin, CWI Amsterdam,
25. Jan. 08, 14 ct, Raum 612

Abtract:
Good distributions of points on the unit sphere in n-dimensional Euclidean space are of interest in pure and applied mathematics, e.g. in coding theory, approximation theory, geometry of numbers, representation theory. Two measures of goodness are maximality and optimality. We say that a subset of N points of the unit sphere is a spherical code with parameters (n,N,t) if any two distinct points have inner product at most t. Then, a maximal spherical code is one which maximizes N given t, and an optimal spherical code is one which minimizes t given N.
Only for very few parameters maximal or optimal spherical codes are known, e.g. finding a maximal (n,N,1/2) spherical code amounts to the kissing number problem which is only known for n = 1, 2, 3, 4, 8, 24.
Traditionally, these problems are treated with elementary geometry and/or with the linear programming bound (LP bound) due to P. Delsarte which generally gives upper bounds for packing problems in two-point homogeneous spaces. Recently, a strengthening of the LP bound, namely the semidefinite programming bound (SDP bound), has been proposed by A. Schrijver in the context of binary codes.
In the talk I will show the general framework behind SDP bounds. In particular I will consider the case of the unit sphere, where we get new upper bounds for the kissing number in dimensions 5, 6, 7, 9, 10, a unified proof of all the cases where the kissing number is known, and a proof that the (4,10,1/6) spherical code, which is closely related to the Petersen graph, is a new optimal spherical code.

This is joint work with Christine Bachoc, based on the papers available from here.


AG Home