🟠 Repeat: L1 vs L2 Speed Round

1. Which drives coefficients to exactly zero?
2. Which is better when features are correlated?
3. Can you apply either to XGBoost?
4. You suspect 90% of your 200 features are noise. Which one?

Answers: 1. L1 (Lasso) 2. L2 (Ridge) — distributes weight among correlated features 3. No — they regularize coefficients, trees don't have coefficients 4. L1 — it will zero out the ~180 irrelevant features