Non-linear Support Vector Machines Explained

With video explanation | Data Series | Episode 9.3

In the previous episode we explained what are support vector machines and the maths behind the algorithm. In this episode we discuss support vector machines for non-linearly separable data.

SVMs for non-linearly separable data

What if the data is not linearly separable? For example:

Calculating a non-linear support vector machine may overfit our data:

In this case we may still want a linear support vector machine but allow it to make some mistakes. This is known as a soft-margin support vector machine.

To do so we make some changes to the SVM objective function defined in the previous episode as :