Statistics for Data Science

Feature space and cheap computations

A huge part of understanding how SVMs work is understanding the trick which we've just mentioned. The trick, as we said, uses kernel methods (which use kernel functions) and are able to perform well in a high-dimensional feature space.

A feature space is an n-dimensional space where your data variables live. Kernel functions are able to operate within this high-dimensional space without having to compute the coordinates of the data within that space, but rather by merely computing the inner products between the images of all pairs of data in the feature space.

This kernel trick can then process the data quicker and more efficiently than if it had to explicitly compute the coordinates. This is known as ...