MLA-C01 Question 149
Select 3You are training a machine learning model using Amazon SageMaker for predicting product demand. During the initial evaluation, the model exhibits high variance, indicating overfitting. Which of the following actions are most appropriate to refine the model and reduce overfitting?
- A
Add dropout layers if using a deep learning model
- B
Increase the size of the training dataset
- C
Reduce the number of features used in the training dataset
- D
Decrease the learning rate used in training
- E
Use L1 or L2 regularization techniques
Show answer and explanation
Correct answers: A, B, E
Explanation
Overfitting occurs when a model performs well on the training data but poorly on unseen data. To address overfitting, strategies such as adding dropout layers, increasing the size of the training dataset, and using regularization techniques like L1 or L2 can help. These methods encourage the model to generalize better by reducing its complexity or exposing it to more diverse data. Options like reducing features and decreasing the learning rate are less direct or effective in handling overfitting.
- A. Correct.
Adding dropout layers is an effective technique to reduce overfitting in deep learning models by randomly deactivating neurons during training, which prevents the model from becoming overly reliant on specific neurons.
- B. Correct.
Increasing the size of the training dataset can help the model generalize better by exposing it to more diverse examples, which reduces overfitting.
- C. Incorrect.
Reducing the number of features can help simplify the model, but it may also result in loss of important information, which is not ideal in all cases for addressing overfitting.
- D. Incorrect.
Decreasing the learning rate affects the convergence speed of the model but does not directly address overfitting.
- E. Correct.
Using L1 or L2 regularization techniques adds penalties to the loss function, discouraging overly complex models and helping reduce overfitting.