1: \begin{abstract}

2: Let $\Omega$ be a bounded closed

3: %\textcolor{blue}{

4: convex set in $\R^d$ with non-empty interior, and let $\cC_r(\Omega)$

5: be the class of convex functions on $\Omega$ with $L^r$-norm bounded by 1.

6: We obtain sharp estimates of the

7: $\eps$-entropy of $\cC_r(\Omega)$ under $L^p(\Omega)$ metrics, $1\le p<r\le \infty$.

8: In particular, the results imply that the universal lower bound $\eps^{-d/2}$ is attained by all $d$-polytopes.

9: While a general convex body can be approximated

10: by inscribed polytopes, the entropy rate does not carry

11: over to the limiting body. For example, if $\Omega$ is the closed unit ball in

12: $\R^d$, then the metric entropy of $\cC_\infty(\Omega)$ under the $L^p(\Omega)$

13: metrics has order $\eps^{-(d-1)p/2}$ for $p>\frac{d}{d-1}$.

14: Our results have

15: applications to questions concerning rates of convergence of nonparametric estimators

16: of high-dimensional shape-constrained functions.

17: \end{abstract}

18: