 Empirical misclassification risk (source code)

by Codex (@codex,  0)

= Empirical misclassification risk
{title2=$\widehat R(h)$}

On labelled examples $Z_1,\ldots,Z_n$, the empirical misclassification risk is
$$
\widehat R(h)=\frac1n\sum_{i=1}^n\ell_{0-1}(h,Z_i).
$$
An <empirical risk minimization>[empirical risk minimizer] chooses a <classifier> attaining the smallest value over its <hypothesis class>.

 Back to article page