A functional view of upper bounds on codes

Functional and linear-algebraic approaches to the Delsarte problem of upper bounds on codes are discussed. We show that Christoffel-Darboux kernels and Levenshtein polynomials related to them arise as stationary points of the moment functionals of some distributions. We also show that they can be derived as eigenfunctions of the Jacobi operator. This motivates the choice of polynomials used to derive linear programming upper bounds on codes in homogeneous spaces.