Which of the following should be used to resolve this issue?
(Choose two.)
Add vanishing gradient to the model.
Perform data augmentation on the training data.
Make the neural network architecture complex.
Use gradient checking in the model.
Add L2 regularization to the model.
Explanations:
Data augmentation helps to artificially increase the diversity of the training dataset, making the model more robust and capable of generalizing better to unseen data. This can reduce overfitting.
L2 regularization adds a penalty to the loss function to discourage the model from becoming too complex, thus preventing overfitting and helping the model generalize better.
Vanishing gradients are a problem, not a solution. They hinder the training of deep networks by making it difficult for the model to learn. Adding vanishing gradient would exacerbate the issue.
Increasing the complexity of the neural network can worsen overfitting, especially when there is insufficient data. This can lead to poor generalization on test data.
Gradient checking is a technique used to verify the correctness of gradients during backpropagation. While useful for debugging, it does not address the issue of overfitting or generalization.
I rank that the answer is:
Perform data augmentation on the training data.
If I had to guess, I’d say the answer is:
Perform data augmentation on the training data.