MLA-C01 Question 48
Select 3You are working on a machine learning project to classify images using Amazon SageMaker. During the training process, you notice that the validation accuracy is significantly lower than the training accuracy. What steps could you take to address this issue and improve the model's performance on the validation set?
- A
Use data augmentation techniques to increase the diversity of the training dataset.
- B
Reduce the size of the training dataset to focus only on the most relevant samples.
- C
Apply regularization techniques, such as L2 regularization or dropout, in the model architecture.
- D
Increase the model's complexity by adding more layers to the neural network.
- E
Perform hyperparameter tuning to find the optimal parameters for the model.
Show answer and explanation
Correct answers: A, C, E
Explanation
The issue of lower validation accuracy compared to training accuracy indicates overfitting. Strategies to address overfitting include increasing training data diversity through augmentation, applying regularization techniques to reduce model complexity, and performing hyperparameter tuning to optimize the model's performance. Reducing the training dataset or increasing model complexity would not address the overfitting issue effectively.
- A. Correct.
Using data augmentation increases the diversity of the training dataset, which helps the model generalize better and reduces overfitting, thereby improving validation performance.
- B. Incorrect.
Reducing the size of the training dataset is unlikely to improve the model's performance on the validation set and may lead to underfitting.
- C. Correct.
Regularization techniques like L2 regularization or dropout prevent the model from overfitting to the training data, which can improve validation accuracy.
- D. Incorrect.
Increasing the complexity of the model may lead to further overfitting and worsen the performance on the validation set.
- E. Correct.
Hyperparameter tuning can help find an optimal configuration for the model to improve its generalization capability and validation performance.