CODEWITHSUNDEEP
pythonnumpymultidimensional-array
Nuageux
Nuageux

Reputation: 1686

Reduction first dimension of ndarray - keep first and last elements

Let's say I have a 3D matrix I want to reduce. I would like to reduce the first dimension only to around 10 elements (it is best if it is 10 but not necessary). I used this code for that:

import numpy
m = numpy.random.rand(37,2,100)
new_m = m[0:-1:int(m.shape[0]/10)]
# new_m.shape = (12, 2, 100)

My problem is the following: I want the first and last elements of this first dimension to be the same than my original matrix. But with my current code, this condition does not hold for the last element:

new_m[0,:,:2], new_m[-1,:,:2]
#[[ 0.06081972  0.91343839] [ 0.89614534  0.33846807]]
#[[ 0.37289341  0.62491196] [ 0.30603305  0.1442681 ]]
m[0,:,:2], m[-1,:,:2]
#[[ 0.06081972  0.91343839] [ 0.89614534  0.33846807]]
#[[ 0.28143018  0.10626664] [ 0.30334235  0.29616713]]

As you can see new_m[-1] is not equal to m[-1], and I would like them to be.

My tries/ideas:

Any ideas, suggestions?

Upvotes: 2

Views: 82

Answers (1)

Arya McCarthy
Arya McCarthy

Reputation: 8829

Since you're using numpy, you can exploit fancy indexing. We'll also use linespace, which generates evenly spaced points in a range, including endpoints. The arguments are similar to what you'd pass to range or your indexing method.

indexes = np.linspace(0, len(m) - 1, 10, dtype=int)
new_m = m[indexes]

Upvotes: 2

Related Questions