Optimal experimental design and quadratic optimization

Rebecca Haycroft (1) , Luc Pronzato (2) , Henry P. Wynn (3) , Anatoly A. Zhigljavsky (1)

1 School of Mathematics [Cardiff]
2 Laboratoire d'Informatique, Signaux, et Systèmes de Sophia-Antipolis (I3S) / Equipe SYSTEMES
3 LSE - Department of Statistics - London School of Economics

A well known gradient-type algorithm for solving quadratic optimization problems is the method of Steepest Descent. Here the Steepest Descent algorithm is generalized to a broader family of gradient algorithms, where the step-length is chosen in accordance with a particular procedure. The asymptotic rate of convergence of this family is studied. To facilitate the investigation, we re-write the algorithms in a normalized form which enables us to exploit a link with theory of optimum experimental design.

Mots clés: gradient algorithms; steepest descent algorithm; rate of convergence; design of experiments; optimality criteria

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Rebecca Haycroft, Luc Pronzato, Henry P. Wynn, Anatoly A. Zhigljavsky. Optimal experimental design and quadratic optimization. ProbaStat 2006, Jun 2006, Smolenice, Slovakia. pp.115-123. ⟨hal-00322795⟩