English

An algorithm for semi-infinite polynomial optimization

Optimization and Control 2011-01-24 v1

Abstract

We consider the semi-infinite optimization problem: f:=minxX{f(x):g(x,y)0,\forallyYx}f^*:=\min_{x\in X}\:\{f(x): g(x,y)\,\leq \,0,\:\forally\in Y_x\}, where f,gf,g are polynomials and XRnX\subset R^n as well as Y\xRpY_\x\subset R^p, xXx\in X, are compact basic semi-algebraic sets. To approximate ff^* we proceed in two steps. First, we use the "joint+marginal" approach of the author to approximate from above the function xΦ(x)=sup{g(x,y):yYx}x\mapsto\Phi(x)=\sup \{g(x,y): y\in Y_x\} by a polynomial ΦdΦ\Phi_d\geq\Phi, of degree at most 2d2d, with the strong property that Φd\Phi_d converges to Φ\Phi for the L1L_1-norm, as dd\to\infty (and in particular, almost uniformly for some subsequence (d)(d_\ell), N\ell\in\N). Then we solve the polynomial optimization problem fd=minxX{f(x):Φd(x)0}f^*_d=\min_{x\in X} \{f(x): \Phi_d(x)\leq0\} via a (by now standard) hierarchy of semidefinite relaxations. It turns out that the optimal value fdff^*_d\geq f^* converges to ff^* as dd\to\infty. In practice we let dd be fixed, small, and relax the constraint Φd0\Phi_d\leq0 to Φd(x)ϵ\Phi_d(x)\leq\epsilon with ϵ>0\epsilon>0, allowing to change ϵ\epsilon dynamically.

Keywords

Cite

@article{arxiv.1101.4122,
  title  = {An algorithm for semi-infinite polynomial optimization},
  author = {Jean Lasserre},
  journal= {arXiv preprint arXiv:1101.4122},
  year   = {2011}
}

Comments

To appear in TO (Spanish Journal of Statistics and Operations Research)

R2 v1 2026-06-21T17:14:59.388Z