What are the training data points that lie closest to the classification boundary called?

What are the training data points that lie closest to the classification boundary called?

• A. Support Vectors
• B. Margin Points
• C. Nearest Points
• D. Boundary Points

Answer: (A)

In a support vector machine, the training examples that sit closest to the decision boundary are called support vectors. They are critical because the boundary is placed to maximize the margin—the gap between the boundary and these nearest points. The position of the separating hyperplane is determined by these points: they effectively define where the margin lies, while all other data lie further away and don’t change the boundary once the support vectors are fixed. In hard-margin SVM, these points sit exactly on the margin lines; in soft-margin, some may be inside the margin or even misclassified, but the ones with the strongest influence (the ones with nonzero Lagrange multipliers) remain the support vectors and continue to shape the boundary.

In a support vector machine, the training examples that sit closest to the decision boundary are called support vectors. They are critical because the boundary is placed to maximize the margin—the gap between the boundary and these nearest points. The position of the separating hyperplane is determined by these points: they effectively define where the margin lies, while all other data lie further away and don’t change the boundary once the support vectors are fixed. In hard-margin SVM, these points sit exactly on the margin lines; in soft-margin, some may be inside the margin or even misclassified, but the ones with the strongest influence (the ones with nonzero Lagrange multipliers) remain the support vectors and continue to shape the boundary.