Huanfa Chen - huanfa.chen@ucl.ac.uk
09/03/2026
Note
This classification is different from unsupervised vs supervised machine learning.

| OLS | Lasso | |
|---|---|---|
| Formula | \(y = X\beta\) | \(y = X\beta\) |
| To minimise | \(\min_{\beta} \left\{ \frac{1}{2n} \sum_{i=1}^{n} (y_i - \mathbf{x}_i^T \beta)^2 \right\}\) | \(\min_{\beta} \left\{ \frac{1}{2n} \sum_{i=1}^{n} (y_i - \mathbf{x}_i^T \beta)^2 + \alpha \|\beta\|_1 \right\}\) |
| VIF | Lasso | |
|---|---|---|
| Purpose | Addresses multicollinearity | Feature selection, including address multicollinearity |
| Type | Unsupervised | Supervised |
| When | Before linear regression | Embedded in linear regression |
| Output | Selected features | Features with non-zero coefficients |
| Model Type | Linear regression | Linear regression |

sklearn): 20,640 houses in California, each described by 8 numeric features (e.g., median income, house age, latitude)
coef_ or feature_importances_ attribute
© CASA | ucl.ac.uk/bartlett/casa