Question 8
Domain 2: Core Machine Learning, AI, and Transformer FoundationsWhat is the purpose of dropout in neural networks?
Correct answer: B
Explanation
Dropout is a regularization method used during training to reduce overfitting. It works by "randomly setting neurons to zero," which prevents the network from relying too heavily on any one feature or pathway and improves generalization.
Why each option is right or wrong
A. To speed up training
B. To prevent overfitting by randomly setting neurons to zero
During training, dropout is applied as a regularization technique: each unit is independently omitted with a specified probability, so the model cannot depend on any single activation pattern and is forced to learn more robust features. In practice, this reduces co-adaptation of neurons and lowers the risk of overfitting; at inference time, the full network is used with the activations scaled appropriately to preserve expected output.
C. To reduce model size
D. To improve interpretability