Regularization (Ridge, Lasso) in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Ridge vs Lasso Regularization

Which statement correctly describes the main difference between Ridge and Lasso regularization?

A Lasso regularization always performs better than Ridge because it uses both L1 and L2 penalties.

B Ridge adds an L1 penalty that sets coefficients to zero, while Lasso adds an L2 penalty that only shrinks coefficients.

C Both Ridge and Lasso add L2 penalties but differ in how they scale the penalty term.

D Ridge adds an L2 penalty which shrinks coefficients but does not set any to zero, while Lasso adds an L1 penalty that can set some coefficients exactly to zero.

❓ Predict Output

intermediate

Output of Lasso Regression Coefficients

What will be the output coefficients after fitting Lasso regression on the given data?

ML Python

from sklearn.linear_model import Lasso
import numpy as np

X = np.array([[1, 2], [2, 3], [3, 4], [4, 5]])
y = np.array([3, 5, 7, 9])

model = Lasso(alpha=1.0)
model.fit(X, y)
coefficients = model.coef_
print(coefficients)

A [0. 0.]

B [1. 1.]

C [0. 2.]

D [0.5 1.5]

❓ Hyperparameter

advanced

Effect of Increasing Alpha in Ridge Regression

What happens to the coefficients of a Ridge regression model as the regularization parameter alpha increases?

A Coefficients remain unchanged regardless of alpha value.

B Coefficients shrink towards zero, reducing model complexity and overfitting.

C Coefficients become larger in magnitude to fit the training data better.

D Coefficients become exactly zero, performing feature selection.

❓ Metrics

advanced

Choosing Between Ridge and Lasso Using Cross-Validation

You have trained both Ridge and Lasso regression models on the same dataset using cross-validation. The Ridge model has a mean squared error (MSE) of 4.5, and the Lasso model has an MSE of 4.7. Which model should you choose and why?

A Choose Ridge because it has a lower MSE, indicating better predictive performance.

B Choose Lasso because it performs feature selection even if MSE is higher.

C Choose Lasso because it always generalizes better than Ridge.

D Choose Ridge only if the number of features is greater than the number of samples.

🔧 Debug

expert

Debugging Lasso Regression with Unexpected Zero Coefficients

You trained a Lasso regression model on your dataset, but all coefficients are zero. What is the most likely cause?

A The alpha parameter is set too high, causing all coefficients to shrink to zero.

B The dataset has too many features, causing Lasso to fail.

C The model was trained without fitting the data.

D The target variable is constant, so coefficients are zero.