Matthias Seeger
Here, I review facts that are most probably known, namely that the information gain criterion used to drive experimental design in a linear-Gaussian model is submodular, so that a well-known approximation guarantee holds for the sequential greedy algorithm ...
2009