What is multiclass classification?

• A. A classification problem with more than two mutually exclusive classes.
• B. A regression task predicting continuous values.
• C. A binary classification with two classes.
• D. A clustering technique grouping similar instances.

Answer: (A)

Multiclass classification is a classification task with more than two mutually exclusive categories, where each instance is assigned exactly one category. This means the output is a single label chosen from several possible classes, unlike regression which predicts a continuous value or binary classification which has only two possible labels. It also differs from clustering, which groups data without predefined labels or outcomes. A practical example is labeling images of handwritten digits 0 through 9, where each image is given one digit as its category. In practice, you can tackle multiclass problems with approaches like one-vs-rest, where separate binary classifiers are trained against each class, or with models that produce a probability distribution over all classes using a softmax layer.