Rates of convergence in sequence-compound squared-distance loss estimation and two-action problems