Expected prediction error on unseen data decomposes as
Two candidate models are trained on the same data-generating process, whose irreducible noise variance is for both.
| Model | Bias | Variance |
|---|---|---|
| A | ||
| B |
The table reports the bias itself, not its square.
Which model has the lower expected MSE, and what is that MSE?
Once you get it right, the reasoning behind the answer shows up here.
Select all that apply.