Which method keeps all features but shrinks coefficients toward zero and penalizes large coefficients to stabilize estimates?

Which method keeps all features but shrinks coefficients toward zero and penalizes large coefficients to stabilize estimates?

• A. Ridge
• B. LASSO
• C. Cook's Distance
• D. Residual Plots

Answer: (A)

Ridge regression uses an L2 penalty on the coefficients, which pulls them toward zero without dropping any features. This shrinkage reduces model variance and stabilizes estimates, especially when predictors are correlated or the data are noisy. Because the penalty doesn’t force coefficients to be exactly zero, all features stay in the model with smaller weights. By contrast, LASSO applies an L1 penalty and can drive some coefficients to exactly zero, effectively removing those features. Cook's Distance and Residual Plots are diagnostic tools for assessing influence and fit, not methods for shrinking coefficients.

Ridge regression uses an L2 penalty on the coefficients, which pulls them toward zero without dropping any features. This shrinkage reduces model variance and stabilizes estimates, especially when predictors are correlated or the data are noisy. Because the penalty doesn’t force coefficients to be exactly zero, all features stay in the model with smaller weights.

By contrast, LASSO applies an L1 penalty and can drive some coefficients to exactly zero, effectively removing those features. Cook's Distance and Residual Plots are diagnostic tools for assessing influence and fit, not methods for shrinking coefficients.