Can some one please explain me what is the difference between one class SVM and SVDD(support vector data description)

  • $\begingroup$ Is there experience around in which cases one of these models is superior? $\endgroup$ – MaxS Dec 20 '18 at 6:16

Support vector data description (SVDD) finds the smallest hypersphere that contains all samples, except for some outliers. One-class SVM (OC-SVM) separates the inliers from the outliers by finding a hyperplane of maximal distance from the origin.

If the kernel function has the property that $k(\mathbf{x}, \mathbf{x}) = 1 \quad \forall \mathbf{x} \in \mathbb{R}^d$, SVDD and OC-SVM learn identical decision functions. Many common kernels have this property, such as RBF, Laplacian and $\chi^2$.

SVDD and OC-SVM are also equivalent in the case that all samples lie on a hypersphere centered at the origin, and are are linearly separable from it.

See Lampert, C. H. (2009). Kernel methods in computer vision (Chapter 5) for more detailed descriptions of these models.

| improve this answer | |
  • $\begingroup$ if we had to solve the svdd as an optimization problem how can we generate the constraints $\endgroup$ – ou2105 Nov 8 '17 at 8:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.