5주차-Vectorization, Locally weighted regression

• Vectorization

 

• Properties of Transpose

 

• Normal Equation

 

• Locally weighted regression

• "Parametic" learning algorithm

Fit fixed set of parameters(w) to data
EX) linear regression, Neural Network

주어진 학습데이터로 파라미터 학습을 완료한 이후에 test를 수행한다.

• "Non-parametic" learning algorithm

Amount of data/parameters you need to keep grows(linearly) with size of data
EX) K-nearest neighbors, Locally weighted regression

test를 수행할 때 학습데이터를 활용

 

• Linear regression

build model f(x) and fit w to minimize

• Locally weighted regression

Fit w to minimize

 

• Locally weighted regression with Γ