Dropout and Stochastic Regularization Quiz

Answers and explanations appear as you take the quiz.

1. What does dropout do during training?
   • Deletes weights permanently
   • Randomly sets a fraction of neuron outputs to zero on each forward pass
   • Doubles the learning rate
   • Adds more layers to the network
2. Why does dropout improve generalization?
   • It increases the number of parameters
   • It removes the need for training data
   • It guarantees zero training error
   • It prevents neurons from co-adapting and forces redundant, robust representations
3. What happens to dropout at test (inference) time?
   • It is turned off and the full network is used
   • It drops even more neurons
   • It deletes half the weights
   • It randomizes the labels
4. A dropout rate of p = 0.5 in a layer means roughly what during training?
   • All neurons are removed
   • No neurons are affected
   • About half the neurons are zeroed out each pass
   • The layer is deleted
5. What is dropout often described as approximating?
   • A single linear regression
   • A decision tree
   • A Fourier transform
   • An ensemble of many thinned networks that share weights
6. What is Monte Carlo dropout used for?
   • Speeding up data loading
   • Estimating model uncertainty by keeping dropout active at inference
   • Compressing the network to fewer weights
   • Labeling the training data automatically
7. With inverted dropout, why is no extra activation scaling normally needed at ordinary inference?
   • The keep probability becomes exactly zero
   • Inference always averages new random masks
   • Surviving activations were already scaled during training
   • Every weight is divided by the number of layers
8. The same input yields varying predictions because dropout remains active during evaluation. What explains the variation?
   • New random masks sample different subnetworks on each pass
   • The input changes its label by itself
   • The network has no trained weights
   • Batch size must be one