NCA-GENM Question 19
Select 3You are tasked with training a multimodal AI model that processes images and text to classify products into categories. During the development process, you notice that the model's performance on the test set is significantly worse than on the training set. To address this issue, which of the following steps would be most appropriate?
- A
Perform cross-validation to ensure the model generalizes well to unseen data.
- B
Add more layers and parameters to the model to increase its capacity.
- C
Use feature engineering to preprocess the text and image data more effectively.
- D
Reduce the size of the training dataset to prevent overfitting.
- E
Compare the performance of different models to identify the best architecture.
Show answer and explanation
Correct answers: A, C, E
Explanation
The problem described indicates overfitting, where the model performs well on the training data but poorly on the test data. Cross-validation (Option 1) is a key technique to evaluate generalization, while feature engineering (Option 3) can help improve data quality for better learning. Additionally, model comparison (Option 5) allows you to identify architectures that might generalize better. Adding more layers (Option 2) and reducing the dataset size (Option 4) are not appropriate solutions to address overfitting.
- A. Correct.
Performing cross-validation helps assess the model's ability to generalize to unseen data by evaluating it on multiple subsets of the data. This can provide insights into whether overfitting or underfitting is occurring.
- B. Incorrect.
Adding more layers and parameters can increase the model's capacity, but this is likely to exacerbate overfitting rather than resolve it.
- C. Correct.
Effective feature engineering, such as preprocessing text and image data appropriately, can improve model performance by ensuring the input features are optimized for learning.
- D. Incorrect.
Reducing the size of the training dataset is counterproductive as it would likely increase overfitting and reduce the model's ability to generalize.
- E. Correct.
Comparing different models helps identify architectures that might be better suited to the dataset, potentially improving generalization and overall performance.