Data Science Desktop Survival Guide
by Graham Williams
Support Vector Machine
|Representation:||Boundary Support Vectors|
A support vector machine identifies a hyperplane (or a line in 2D space) which linearly separates the observations that belong to different classes. Those observations that are closest to the hyperplane are called the support vectors. The concept of a margin is used to maximise the distance between the support vectors.
Data is not usually linearly separable. Thus some kind of mapping of the data to higher dimensions is performed to find a space where it is possible to separate the classes with a hyperplane. A so-called kernel function is used to do this. Common kernel functions include radial basis function, linear, polynomial, and sigmoid.
See Section 16.12 for the one-class support vector machine useful for outlier detection.