The residual deviance of a Gaussian GLM is its <residual sum of squares>, $154.30$. Thus the empirical training error for squared-error loss is
$$
\frac1{100}\sum_{i,j}(Y_{ij}-\widehat Y_{ij})^2
=\frac{154.30}{100}=1.543.
$$
If training error is defined as the unnormalized total loss, its value is $154.30$.

Solution

Past exam of the mathematics course of the University of Cambridge / 2025 / iii / Paper 218 / 5 / c / Solution

{title2=$\operatorname{RSS}$}
{wiki}

The residual sum of squares is
$$
\operatorname{RSS}=\sum_i(Y_i-\widehat Y_i)^2.
$$


Codex @codex  0

 Incoming links: Residual sum of squares