Supervised Machine Learning with Python

Distance metrics

So, before we get into the algorithms, I want to address some mathematical esotericism. When you have two points, or any number of points, in a 2D space, it's fairly easy to conceptualize. It's basically calculating the hypotenuse along some right triangle, in terms of measuring the distance. However, what happens when you have a really high dimensional space? That's what we're going to get into, and we have a lot of clustering problems.

So, the most common distance metric is the Euclidean distance. This is essentially a generalization of the 2D approach. It's the square root of the sum of squared differences between two vectors, and it can be used in any dimensional space. There's a lot of others that we're not going to ...