Google Professional Machine Learning Engineer Question 244
Select 4Google Cloud PlatformYou are training a machine learning model on Google Cloud using Vertex AI. During the training process, you notice the model is overfitting on the training dataset. Which of the following actions should you take to mitigate overfitting?
- A
Increase the size of the training dataset by incorporating more diverse examples.
- B
Reduce the complexity of the model architecture by using fewer layers or parameters.
- C
Increase the number of training epochs to ensure the model learns the patterns more thoroughly.
- D
Apply regularization techniques such as L1 or L2 regularization.
- E
Use techniques like dropout to randomly deactivate certain neurons during training.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
Overfitting occurs when a model performs well on the training data but poorly on unseen data. To mitigate overfitting, you can take measures such as increasing the training dataset size, simplifying the model architecture, applying regularization techniques like L1/L2 regularization, and using dropout. Increasing training epochs, however, may worsen overfitting by allowing the model to memorize the training data further.
- A. Correct.
Increasing the size of the training dataset helps the model generalize better and reduces the likelihood of overfitting as it learns from more diverse examples.
- B. Correct.
Reducing the complexity of the model architecture can prevent it from memorizing the training data, which is a common cause of overfitting.
- C. Incorrect.
Increasing the number of training epochs may exacerbate overfitting as the model will continue to memorize the training data rather than generalizing.
- D. Correct.
Regularization techniques, like L1 or L2 regularization, add penalties to the loss function, discouraging the model from relying too heavily on specific features and reducing overfitting.
- E. Correct.
Dropout is a widely used regularization technique that prevents overfitting by randomly deactivating neurons during training, forcing the model to generalize better.