A new algorithm for optimal, distance-based experimental design

A new algorithm for optimal, distance based, experimental design is presented. The proposed approach, in contrast to other optimality criteria, does not require any preliminary hypothesis about a regression model. The best set of experiments is defined through a fast exchange algorithm where, in each cycle, a substitution is selected to provide the maximum increase of the minimum distance between the currently selected experiments. Such an algorithm provides a final uniform distribution of the experiments selected from the set of allowed candidates.