MLS-C01 Question 192
Single answerYou are training a machine learning model on AWS SageMaker to predict customer churn. The dataset has a large number of features, some of which are highly correlated and may lead to overfitting. Which regularization technique should you apply to the model to reduce overfitting while ensuring sparsity in feature selection?
- A
L1 regularization
- B
L2 regularization
- C
Dropout regularization
- D
Batch normalization
Show answer and explanation
Correct answer: A
Explanation
L1 regularization is the most suitable choice in this scenario because it not only helps reduce overfitting but also encourages sparsity by penalizing the absolute values of weights. This can drive some weights to zero, effectively selecting a subset of important features. L2 regularization, while useful for reducing overfitting, does not provide the sparsity required in feature selection. Dropout and batch normalization serve different purposes and are not relevant to this problem.
- A. Correct.
L1 regularization (also known as Lasso) is effective at reducing overfitting and encourages sparsity in the model by penalizing the absolute values of weights. This helps in feature selection by driving some weights to zero.
- B. Incorrect.
L2 regularization (also known as Ridge) reduces overfitting by penalizing the squared values of weights, but it does not encourage sparsity. It doesn't drive weights to zero, so it is not ideal for feature selection in this case.
- C. Incorrect.
Dropout regularization is primarily used in neural networks to prevent overfitting by randomly dropping units during training. It does not directly address sparsity or feature selection.
- D. Incorrect.
Batch normalization is a technique to stabilize and speed up training by normalizing inputs to a layer. It is not a regularization method aimed at reducing overfitting or encouraging sparsity.