CODEWITHSUNDEEP
pythonnumpyscikit-learnscipy
Johannes Wiesner
Johannes Wiesner

Reputation: 1307

Compute distance between vector and matrix

I am using sklearn.metrics.pairwise.paired_distances to calculate distances between a single vector and a matrix. I want to calculate the distance between every row of the matrix and the single vector. Since sklearn.metrics.pairwise.paired_distances requires the two arrays to have equal dimensions, I use np.tile to create a matrix which contains multiple copies of the vector to create a matrix that has the same size as the first one.

Example:

import numpy as np
from sklearn.metrics.pairwise import paired_distances

# get matrix a and vector b
a = np.array([[1,2],[3,4]])
b = np.array([[5],[6]]).transpose()

# create a matrix with copies of b that has the same size as matrix a
b = np.tile(b,(a.shape[0],1))

distances = paired_distances(a,b)

Just out of curiosity: Is there a function that does that out of the box? Time is not critical here, since I don't deal with very big arrays. But the function should offer different kinds of metrics.

Upvotes: 0

Views: 467

Answers (1)

mutantacule
mutantacule

Reputation: 7063

You could use numpy.apply_along_axis method to apply a given function to each rows.

Upvotes: 1

Related Questions