True, when there are three or more classes: each input gets exactly one label. With two classes it's binary classification; when an input can carry several labels at once, it's multi-label classification.
Proposed by @svpino
Card 1 of 259. Answer: True, when there are three or more classes: each input gets exactly one label. With two classes it's binary classification; when an input can carry several labels at once, it's multi-label classification.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
True, provided there are more than two classes. Multi-class classification assigns each input exactly one label from three or more mutually exclusive classes, such as which of the 10 digits a handwritten image shows. With two classes it is binary classification; when one input can belong to several classes at once, it is multi-label classification.
The difference changes the output layer. A multi-class network ends in K softmax units trained with categorical cross-entropy and predicts the argmax: scores for cat, dog and bird might be 0.7, 0.2 and 0.1, summing to 1. A multi-label model tagging a photo of a cat and a dog uses K independent sigmoids with binary cross-entropy and thresholds each, so outputs like 0.9, 0.8 and 0.1 need not sum to 1.
A common follow-up is how to apply binary classifiers to multi-class data. One-vs-rest trains K classifiers, one per class; one-vs-one trains one per pair, K(K − 1)/2, so 45 for 10 classes. Per-class metrics are then combined by macro or micro averaging.
