MLS-C01 Question 186
Single answerYou are working on a regression problem and your model is overfitting the training data. To address this, you decide to introduce regularization into your model. Which regularization technique would be most appropriate if you want to reduce the impact of less important features by driving their coefficients to exactly zero?
- A
L1 regularization (Lasso Regression)
- B
L2 regularization (Ridge Regression)
- C
Dropout regularization
- D
Batch normalization
Show answer and explanation
Correct answer: A
Explanation
L1 regularization, also known as Lasso Regression, is designed to reduce overfitting by shrinking some feature coefficients to zero, effectively removing less important features from the model. This makes it a powerful tool for both regularization and feature selection. L2 regularization, while also effective for reducing overfitting, does not perform feature selection as it shrinks coefficients without making them exactly zero. Dropout regularization and batch normalization are techniques specific to neural networks and are not applicable to this regression scenario.
- A. Correct.
L1 regularization (Lasso Regression) works by adding a penalty proportional to the absolute value of the coefficients. This can shrink less important feature coefficients to exactly zero, effectively performing feature selection. This makes it the correct choice for the scenario described.
- B. Incorrect.
L2 regularization (Ridge Regression) adds a penalty proportional to the squared value of the coefficients, which shrinks coefficients but does not drive them to exactly zero. It is not suitable for feature selection.
- C. Incorrect.
Dropout regularization is used to prevent overfitting in neural networks by randomly dropping units during training. While it helps with overfitting, it is not related to the shrinking of feature coefficients in regression.
- D. Incorrect.
Batch normalization is a technique to stabilize and accelerate training in deep learning models by normalizing input layers. It is not a regularization method for regression models and does not address overfitting by modifying feature coefficients.