线性回归

Linear Regression

• It's not just fitting lines and planes!

Setup:

Goal:

Basis fun: represent nonlinear problems

Simple case

Basis func, It is still linear in w

Polynomials:

Radial basis fun : high dim

Fouier basis:

wavelets: image processing

Linear in w, not nec. in x !

The left one is called original space and right one is feature space. sometimes, it is not linear in orignial space but linear in feature space.

---

Discriminative and generative model