[FIXED] Why use only the first column from the distances?

Issue

I have a clustering (Kmeans was applied and clusters were obtained). The radius for each cluster is being calculated between the center and the observations.

I don't understand this here [:, 0] I know we're taking all observations from the first column but why not take the second column as well? What does [:, 0] represent?

X_distances = euclidean_distances(X, [center])[:, 0]
radius = np.max(X_distances)

Solution

Whenever you come across things like this, unpack the code a bit and look at what each piece is giving you.

In this case, the docs for sklearn.metrics.pairwise.euclidean_distances — which I assume you are using (please include this sort of information in your questions!) — also tell you what you need to know:

Returns: distances: ndarray of shape (n_samples_X, n_samples_Y)

So the function euclidean_distances(X, Y) returns a 2D array of all the distances between the points in X and the points in Y. Your X is all your data, and your Y is just one point: the centroid of the cluster. Because Y is only one point, your resulting distance matrix has only one column. Like this:

from sklearn.metrics import euclidean_distances
import numpy as np

X = np.array([[1, 3], [2, 5], [0, 4]])
euclidean_distances(X, [[0, 0]])

This gives:

array([[3.16227766],
       [5.38516481],
       [4.        ]])

So the index [:, 0] is getting this column. In fact, you could skip the indexing because np.max() doesn't care: it's just going to give you the max of the entire array. So you could reduce your code to:

radius = euclidean_distances(X, [center]).max()

Answered By - kwinkunks