Polynomial Regression

Question 1

What does Polynomial Regression allow you to do?

Answer 1

• to fit a curve to your data instead of just a flat line

Question 2

What does Polynomial Regression make use of?

Answer 2

• linear regression
• feature engineering
• quadratic and cubic functions
  -> feature x is raised to quadratic and cubic powers
  f_wb(x) = w1x + w2x2 + w3x3
• square root of x
  f_wb(x) = w1x + w2sqrt(x)

Question 3

How important is feature scaling for polynomial functions?

Answer 3

• of increasing important because of quadratic and cubic powers of the features
  -> because the quadratic and cubic value ranges are very different than the simple feature

Question 4

How can Gradient Descent recognize which features are important and which not?

Answer 4

• assuming you have multiple features and try out if quadratic or cubic engineered featured are useful you add them to the function f_wb(x)
• gradient descent runs and applies weights to each feature
• important features receive higher weights then less important features
  -> thus gradient descent marks whether or not quadratic or cubic features are valuable or not