H13-311_V3.5 · Question #348
What are the regularizations in deep learning? (Multiple Choice)
The correct answer is A. L1 norm. L2 norm B. Data set enhancement C. Integration method D. Dropout. All four options are valid regularization techniques in deep learning. L1/L2 norms (A) add penalty terms to the loss function that shrink weights - L1 promotes sparsity, L2 penalizes large weights uniformly, both reducing overfitting. Data augmentation (B) expands the effective…
Question
What are the regularizations in deep learning? (Multiple Choice)
Options
- AL1 norm. L2 norm
- BData set enhancement
- CIntegration method
- DDropout
How the community answered
(52 responses)- A100% (52)
Explanation
All four options are valid regularization techniques in deep learning. L1/L2 norms (A) add penalty terms to the loss function that shrink weights - L1 promotes sparsity, L2 penalizes large weights uniformly, both reducing overfitting. Data augmentation (B) expands the effective training set through transformations (flipping, cropping, noise), forcing the model to generalize rather than memorize. Ensemble/integration methods (C) combine multiple models, averaging out individual models' overfitting tendencies. Dropout (D) randomly deactivates neurons during training, preventing co-adaptation and acting like implicit ensemble learning.
Since all choices are correct, there are no wrong distractors - the trap here is assuming only "classical" techniques like L1/L2 count as regularization.
Memory tip: Think "LEAD" - L1/L2, Ensemble, Augmentation, Dropout - any technique that reduces a model's ability to overfit the training data is regularization, whether it modifies the loss, the data, or the network itself.
Topics
Community Discussion
No community discussion yet for this question.