CODEWITHSUNDEEP
pythonnumpyvectoreuclidean-distance
fibbi
fibbi

Reputation: 71

Distance between multiple vectors

I have two sets of three-dimensional unit-vectors that I would like to get a measure of how similar they are. My current method is to manually calculate the euclidean norm of their difference.

An example (without considering unit-vectors) would look like

import numpy as np

N = 4000
a = np.random.rand(3,N)
b = np.random.rand(3,N)

dist = np.sum((a-b)**2,axis=0)
dist = np.sum(dist)/len(dist)

Are there better ways to do this; i.e. a numpy function or a measure I haven't thought of?

Upvotes: 1

Views: 3450

Answers (3)

Levi Lutz
Levi Lutz

Reputation: 161

As discussed in the question comments, this should be more appropriate for what you're trying to do:

from scipy import spatial
import numpy as np

N = 10
a = np.random.rand(3, N)
b = np.random.rand(3, N)

mean_a = np.mean(a, axis=1)
mean_b = np.mean(b, axis=1)

similarity = spatial.distance.cosine(mean_a, mean_b)

Values closer to 1 are more similar, values closer to 0 are less similar.

Upvotes: 0

Alexander Golys
Alexander Golys

Reputation: 703

To calculate Euclidean distance in numpy you can use 

numpy.linalg.norm(a-b)

Also there are other types of distances, measuring other types of similarities. SciPy has a lot of them implemented and described in docs: https://docs.scipy.org/doc/scipy/reference/spatial.distance.html

Upvotes: 3

maede rayati
maede rayati

Reputation: 786

You can use distance in scipy package.

from scipy.spatial import distance

distance.euclidean(a, b)

Upvotes: 0

Related Questions