= Solution
The exponential loss is $\phi(z)=e^{-z}$. For a score $f:\mathcal X\to\mathbb R$, its <exponential classification risk> and empirical risk are
$$
R_\phi(f)=\mathbb E[e^{-Yf(X)}],
\qquad
\widehat R_\phi(f)=\frac1n\sum_{i=1}^ne^{-Y_if(X_i)}.
$$
For the set of evaluation vectors
$$
B(x_{1:n})
=\{(h(x_1),\ldots,h(x_n)):h\in B\},
$$
its empirical <Rademacher complexity> is
$$
\boxed{
\widehat R(B(x_{1:n}))
=\frac1n\mathbb E_\sigma
\sup_{h\in B}\sum_{i=1}^n\sigma_ih(x_i)},
$$
where the $\sigma_i$ are independent uniform random signs.
Back to article page
