smote

**SMOTE** (Synthetic Minority Over-sampling Technique) is a **data augmentation method for imbalanced datasets that generates synthetic minority samples by interpolating between existing minority examples** — creating new, diverse training examples along the line segments connecting minority class nearest neighbors. **How SMOTE Works** - **Select**: Choose a minority class sample $x_i$. - **Neighbors**: Find its $k$ nearest minority class neighbors. - **Interpolate**: $x_{new} = x_i + lambda (x_{nn} - x_i)$ where $lambda sim U(0,1)$ and $x_{nn}$ is a random neighbor. - **Repeat**: Generate enough synthetic samples to reach the desired class balance. **Why It Matters** - **Diversity**: Unlike random duplication, SMOTE creates NEW examples — reduces overfitting risk. - **Feature Space**: Interpolation in feature space produces plausible new examples. - **Foundational**: SMOTE (Chawla et al., 2002) is the most cited imbalanced learning method — the standard baseline. **SMOTE** is **creating synthetic minorities** — generating new minority examples by interpolating between existing ones for balanced, diverse training.