multi-label classification

Multi-label classification is a supervised learning task where one example can be assigned several labels at once, rather than being restricted to exactly one class. A news story, for example, might be tagged both “technology” and “business,” while a photo could contain multiple recognized objects. A common approach trains the model to estimate a separate score for each possible label and then applies a threshold or other decision rule to decide which labels to return. This differs from multi-class classification, where the classes are mutually exclusive and a single choice is made. Evaluation needs to account for both correct and missed labels, using measures such as precision, recall, and F1 adapted to the multi-label setting.