Classification Loss and Training

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. Which loss commonly fits one binary target with a predicted probability?A. Binary cross-entropyB. Categorical cross-entropy over ten mutually exclusive classesC. Hinge loss applied directly as a probabilityD. Mean absolute error on class names

2. What does a hinge loss encourage?A. Every prediction to equal zeroB. Every output to sum to oneC. The removal of all decision boundariesD. A sufficient classification margin

3. What is one training epoch?A. Exactly one gradient update in every setupB. The full lifetime of a deployed modelC. One pass through the training datasetD. One prediction on a single validation image

4. A training set has 100 examples and batch size 20, with no dropped examples. How many batches make one epoch?A. 100B. 5C. 20D. 80

5. A three-class classifier predicts 0.1, 0.8, and 0.1. The true class is the second. Which probability is used inside its categorical cross-entropy term?A. 0.8B. 0.1 from the first classC. The largest incorrect probabilityD. The sum of only the incorrect probabilities

Review this module from its first lesson →

Classification Loss and Training

Your result