Universal Regression

Construct an asymptotically optimal prediction algorithm for the problem of regression (i.e., with $\mathbf{Y}=\mathbb{R}$ as the label space) along the lines of Section 3.2 of Vovk et al. 2005 (who in their Theorem 3.1 only consider the case of classification, i.e., finite $\mathbf{Y}$).

Bibliography