What term describes a type of classification in SVMs where all examples are on the correct side of the margin?

The term that describes a type of classification in Support Vector Machines (SVMs) where all examples are on the correct side of the margin is known as hard-margin classification. In this approach, the goal is to find the hyperplane that maximally separates the two classes while ensuring that all data points are classified correctly and are at least a certain distance away from the hyperplane, defined as the margin.

In hard-margin classification, the assumption is made that the data is linearly separable without any outliers, meaning that no data points fall within the margin boundaries. This strict requirement can lead to issues if the data contains noise or cannot be perfectly separated, making hard-margin classification less robust in certain situations.

Soft-margin classification, on the other hand, allows for some misclassifications or points to fall within the margin in order to achieve better generalization on datasets that are not perfectly separable. Thus, while it is a valid classification approach, it does not encapsulate the characteristic of having all examples on the correct side of the margin.

Linear classification and quadratic classification refer to the nature of the decision boundary rather than the separation condition itself. Linear classification uses a linear decision boundary, while quadratic classification allows for a parabolic decision boundary.