AWS Certified Machine Learning Specialty · Free Practice Question Medium
Question 1
The data science team at an analytics company is working on a linear regression model and it observes that the training error as well as the test error are high, implying that the model has a bias.
Which of the following L1 and L2 regularization optimizations may be done to resolve this issue? (Select two)
-
A
Decrease L1 regularization
-
B
Increase L1 regularization
-
C
Use L2 regularization and drop L1 regularization
-
D
L1 and L2 regularization are not required, just get more training data
Reveal correct answers
Correct answers: A, C
Explanation
Correct options:
Use L2 regularization and drop L1 regularization
Decrease L1 regularization
You can think of L1 as reducing the number of features in the model altogether. Think of “Decreasing L1 regularization” as keeping more features in the model, thereby removing the bias.
L2 “regulates” the feature weight instead of just dropping them, so you can use L2 regularization to address the bias.
Highly recommend to review L1 and L2 regularizations in more detail here:
https://towardsdatascience.com/l1-and-l2-regularization-methods-ce25e7fc831c
Incorrect options:
Increase L1 regularization - This option contradicts the explanation provided above, so this option is incorrect.
L1 and L2 regularization are not required, just get more training data - Getting more training data alone will not address the model's bias, so this option is incorrect.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
