Google Professional Machine Learning Engineer Question 247
Select 3Google Cloud PlatformYou are training a machine learning model using TensorFlow on Google Cloud AI Platform (Vertex AI). During training, you notice that the model's accuracy on the validation set is significantly worse than its accuracy on the training set, and the gap increases as training progresses. Which of the following actions should you take to address this issue?
- A
Increase the size of the training dataset by augmenting the data or collecting more samples.
- B
Use regularization techniques such as L1 or L2 regularization in your model.
- C
Reduce the number of features used in the input data to simplify the model.
- D
Switch to a more complex model architecture to better capture the patterns in the data.
- E
Monitor the training process and implement early stopping to prevent overfitting.
Show answer and explanation
Correct answers: A, B, E
Explanation
The problem described is a classic case of overfitting, where the model performs well on training data but struggles on validation data. Effective ways to address overfitting include increasing the training dataset size, using regularization to constrain the model's complexity, and implementing early stopping to halt training before the model overfits. Simplifying the model input (e.g., reducing features) or increasing model complexity are less effective or counterproductive in this context.
- A. Correct.
Increasing the size of the training dataset can help reduce overfitting by providing the model with more diverse data to learn from, reducing its reliance on patterns specific to the training set.
- B. Correct.
Regularization techniques such as L1 or L2 penalize large weights in the model, which can help prevent overfitting by constraining the model's complexity.
- C. Incorrect.
Reducing the number of features may not directly address overfitting unless the features are noisy or irrelevant. Instead, focusing on regularization or increasing the dataset size is more effective.
- D. Incorrect.
Switching to a more complex model architecture can exacerbate overfitting by making the model even more prone to capturing noise or irrelevant patterns in the training data.
- E. Correct.
Early stopping is an effective technique to mitigate overfitting by halting training when the validation performance stops improving, preventing the model from over-optimizing on the training data.