Generalization and Regularization

Five questions. Choose one answer for each, then submit to see your score and explanations.

The practice target is 80%. You can retry as often as you like, and every lesson stays accessible.

You can practice as a guest. Sign in before starting an attempt to save your scores.

1. What is generalization?A. Remembering every training label exactlyB. Increasing model size indefinitelyC. Keeping training loss above validation lossD. Performing well on relevant unseen examples

2. How is standard dropout normally used?A. Keep dropping all units until loss reaches zeroB. Replace the validation set with training dataC. Randomly drop activations during training; disable that randomness for ordinary inferenceD. Randomly delete saved labels during inference

3. Which regularizer tends to encourage sparse weights?A. A larger input image by itselfB. L1 regularizationC. L2 regularization in all casesD. Unregularized training

4. Training loss keeps falling while validation loss rises. Which response is most appropriate?A. Try early stopping or stronger regularization and compare validation resultsB. Always train longer because training loss improvedC. Tune directly on the final test setD. Remove every validation example

5. You horizontally flip a training image for object detection. What must happen to its bounding boxes?A. Leave coordinates unchangedB. Delete all boxes for that imageC. Use coordinates from an unrelated imageD. Transform the box coordinates to match the flipped image

Review this module from its first lesson →

Generalization and Regularization

Your result