Lasso Regression

Linear regression with L1 regularization that drives some coefficients to exactly zero (automatic feature selection).

When to use:

• Have many irrelevant features
• Want automatic feature selection
• Need sparse models (many coefficients = 0)
• Interpretability important

Strengths: Automatic feature selection, creates sparse models, interpretable, prevents overfitting Weaknesses: Drops correlated features arbitrarily, may underperform with many relevant features

Model Parameters

Alpha (default: 1.0) Regularization strength. Higher = more features zeroed out.

• Low (0.01-0.1): Weak selection
• Default (1.0): Moderate selection
• High (10+): Aggressive selection, very few features

Max Iterations (default: 1000) Maximum number of iterations for optimization.

Fit Intercept (default: true) Whether to calculate intercept term.

Selection

• cyclic: Loop through features in order (default)
• random: Random feature selection (can be faster)