Support Vector Data Description

lastmod 2021-06-25created 2021-06-251 edits

About

• What is SVDD? - SVDD (Support Vector Data Description)

SVDD (Support Vector Data Description) is a kind of Boundary-based Novelty Detection, an algorithm developed by Tax and Duin in 2004. Its use is similar to 1-SVM in that it provides a boundary for performing Binary classification, but whereas 1-SVM aims to find the hyperplane boundary farthest from the origin, SVDD can be described as an algorithm whose objective function is to find the minimum-volume sphere enclosing the normal data, providing a boundary for outlier detection.

See also

• SVM
• Binary classification
• Anomaly detection
• Patch SVDD