Answer-first summary for fast verification
Answer: Add L1/L2 regularization., Use training data augmentation.
The question asks for two actions to reduce overfitting in a CNN for image classification. Overfitting occurs when a model learns the training data too well, including noise, and fails to generalize to new data. Option B (Add L1/L2 regularization) is correct because regularization adds penalty terms to the loss function, discouraging complex weights and reducing overfitting. Option C (Use training data augmentation) is correct as it artificially expands the training dataset with transformations (e.g., rotations, flips), improving generalization. The community discussion strongly supports B and C, with high upvotes and consensus (e.g., 100% agreement in answers_community). Options A and E (adding dense layers) increase model complexity, potentially worsening overfitting. Option D (reducing training data) would likely cause underfitting, not reduce overfitting.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You are building a deep convolutional neural network (CNN) for image classification. The model shows signs of overfitting. You need to reduce overfitting and converge the model to an optimal fit.
Which two actions should you perform? Each correct answer presents a complete solution.
NOTE: Each correct selection is worth one point.
A
Add an additional dense layer with 512 input units.
B
Add L1/L2 regularization.
C
Use training data augmentation.
D
Reduce the amount of training data.
E
Add an additional dense layer with 64 input units.
No comments yet.