Support Vector Machines

• SVMs combine support vectors and the kernel trick.
• It finds the support vectors to calculate the maximum margin separators.
• If it can't find a line to separate the classes, it uses the kernel trick to project to higher dimensions where it can find a linear (hyper-planer) separator.
• This is a really popular technique.
• You might need lots of lines to break up the classes.
• In fact, there is no limit.
• So, is this parametric or non-parametric?
• Also note that SVMs aren't just a single algorithm; it's more a set of tools.
• That's good for analytics.