MLA-C01 Question 223
Select 3You are training a machine learning model for a regression task using Amazon SageMaker. During evaluation, you notice the model performs well on the training dataset but poorly on the validation dataset. Which of the following methods can help you identify whether the model is overfitting or underfitting?
- A
Analyze the learning curve to compare training loss and validation loss over time
- B
Increase the complexity of the model by adding more layers or parameters
- C
Check for high variance in model predictions across different validation folds
- D
Reduce the size of the training dataset and observe model performance
- E
Evaluate the model's performance on a test dataset that was not used during training
Show answer and explanation
Correct answers: A, C, E
Explanation
To determine whether a model is overfitting or underfitting, it is crucial to compare training and validation performance. Learning curves can reveal patterns of overfitting or underfitting, high variance in validation predictions can indicate overfitting, and test dataset evaluation ensures that the model generalizes well to unseen data. Methods like increasing model complexity or reducing the training set size do not directly help identify these issues.
- A. Correct.
Analyzing the learning curve is a key method to identify overfitting or underfitting. If the training loss is much lower than the validation loss, it indicates overfitting. If both losses are high, it suggests underfitting.
- B. Incorrect.
Increasing the model complexity can exacerbate overfitting but does not help in identifying whether the model is overfitting or underfitting.
- C. Correct.
High variance in predictions across validation folds often indicates overfitting since the model fails to generalize well across different subsets of data.
- D. Incorrect.
Reducing the size of the training dataset is unlikely to provide insights into overfitting or underfitting; it may degrade model performance instead.
- E. Correct.
Evaluating the model on a test dataset can help confirm whether the model is overfitting if its performance degrades on unseen data.