English

Integrating products of quadratic forms

Probability 2020-02-19 v1 Data Structures and Algorithms Metric Geometry Optimization and Control

Abstract

We prove that if q1,,qm:RnRq_1, \ldots, q_m: {\Bbb R}^n \longrightarrow {\Bbb R} are quadratic forms in variables x1,,xnx_1, \ldots, x_n such that each qkq_k depends on at most rr variables and each qkq_k has common variables with at most rr other forms, then the average value of the product (1+q1)(1+qm)\left(1+ q_1\right) \cdots \left(1+q_m\right) with respect to the standard Gaussian measure in Rn{\Bbb R}^n can be approximated within relative error ϵ>0\epsilon >0 in quasi-polynomial nO(1)mO(lnmlnϵ)n^{O(1)} m^{O(\ln m -\ln \epsilon)} time, provided qk(x)γx2/r|q_k(x)| \leq \gamma \|x\|^2 /r for some absolute constant γ>0\gamma > 0 and k=1,,mk=1, \ldots, m. When qkq_k are interpreted as pairwise squared distances for configurations of points in Euclidean space, the average can be interpreted as the partition function of systems of particles with mollified logarithmic potentials. We sketch a possible application to testing the feasibility of systems of real quadratic equations.

Keywords

Cite

@article{arxiv.2002.07249,
  title  = {Integrating products of quadratic forms},
  author = {Alexander Barvinok},
  journal= {arXiv preprint arXiv:2002.07249},
  year   = {2020}
}

Comments

15 pages

R2 v1 2026-06-23T13:44:37.247Z