Parametric Methods to Estimate $f$

These models have the potential to fit $f$ poorly compared to non-parametric methods, but they are easier to interpret for both inference and prediction purposes.

Make an assumption about the functional form of $f$. Such as a linear form:

Use the training data to estimate the parameters ${\beta }_0,{\beta }_1,\cdots ,{\beta }_p$ such that

Sources: 1