Optimum experimental designs with a lack of a priori information. I. Designs for the estimation of a finite-dimensional set of functionals.