How to determine z for logistic regressions

[agi1]

• Link to the classroom item: https://www.coursera.org/learn/machine-learning/lecture/erGPe/the-problem-of-overfitting
• Description: “The problem of overfitting”

I understand that having squared terms in z generalizes the model well, but i am confused about how we decide z? is it just random like choose whatever you want?

[TMosh]

Start with the simplest linear combination (the figure on the left).

If you don’t get good enough performance, then you start adding more quadratic terms.