Regularization

Regularization

1. Bias & Variance in ML

   • Goal of ML: fit data with right balance between complex and simple

2. OLS: the sum of squared residual is the cost function--->minimize

3. Ridge Regression: Loss func + L2 norm for Penality

   • Works by shrink the parameters of some variable

4. Lasso Regression: Loss func + L1 norm

   • works by remove some of the variable completely

5. Geometry of Lasso & Ridge

6. Elastic Net: combine both penalty term

7. other regularization method