[머신러닝] Introduction

해당 글은 교재 Statistical Learning with R을 챕터별로 정리한 글입니다.

statistical learning : understanding data with vast tools

 

Supervised Learning vs Unsupervised Learning

• supervised : given input and output ex) regression(wage), classification(stock)
• unsupervised : Given only input ex) dimension reduction, clustering

 

Prediction vs Inference

• prediction : predict Y from X
• inference : understand relationship -> want to know what really make f

 

Reducible vs Irreducible errors

• reducible : able to reduce with updating f
• irreducible : only god knows

 

MSE (Mean Squared Error)

• E(Y - Y hat)
• estimator of f
• hope reducible error to be minimized

 

Parametic vs Non-Parametic method

• Parametic : assumption of model shape ex) linear
• Non-parametic : Not assumption

 

Flexibility vs Interpretability

• trade-off
• flexible -> overfitting
• interpretability -> underfitting

 

Test MSE vs Training MSE

• U-shape
• can be different outcome
• too flex -> overfitting

 

Bias-Variance trade-off

• Bias : lower flexible, may closer to True f, more K in KNN
• Variance : more flexible with training data, less K in KNN