Can numpy diagonalise a skew-symmetric matrix with real arithmetic? - python

Any skew-symmetric matrix (A^T = -A) can be turned into a Hermitian matrix (iA) and diagonalised with complex numbers. But it is also possible to bring it into block-diagonal form with a special orthogonal transformation and find its eigevalues using only real arithmetic. Is this implemented anywhere in numpy?


Let's take a look at the dgeev() function of the LAPACK librarie. This routine computes the eigenvalues of any real double-precison square matrix. Moreover, this routine is right behind the python function numpy.linalg.eigvals() of the numpy library.
The method used by dgeev() is described in the documentation of LAPACK. It requires the reduction of the matrix A to its real Schur form S.
Any real square matrix A can be expressed as:
A=QSQ^t
where:
Q is a real orthogonal matrix: QQ^t=I
S is a real block upper triangular matrix. The blocks on the diagonal of S are of size 1×1 or 2×2.
Indeed, if A is skew-symmetric, this decomposition seems really close to a block diagonal form obtained by a special orthogonal transformation of A. Moreover, it is really to see that the Schur form S of the skew symmetric matrix A is ... skew-symmetric !
Indeed, let's compute the transpose of S:
S^t=(Q^tAQ)^t
S^t=Q^t(Q^tA)^t
S^t=Q^tA^tQ
S^t=Q^t(-A)Q
S^t=-Q^tAQ
S^t=-S
Hence, if Q is special orthogonal (det(Q)=1), S is a block diagonal form obtained by a special orthogonal transformation. Else, a special orthogonal matrix P can be computed by permuting the first two columns of Q and another Schur form Sd of the matrix A is obtained by changing the sign of S_{12} and S_{21}. Indeed, A=PSdP^t. Then, Sd is a block diagonal form of A obtained by a special orthogonal transformation.
In the end, even if numpy.linalg.eigvals() applied to a real matrix returns complex numbers, there is little complex computation involved in the process !
If you just want to compute the real Schur form, use the function scipy.linalg.schur() with argument output='real'.
Just a piece of code to check that:
import numpy as np
import scipy.linalg as la
a=np.random.rand(4,4)
a=a-np.transpose(a)
print "a= "
print a
#eigenvalue
w, v =np.linalg.eig(a)
print "eigenvalue "
print w
print "eigenvector "
print v
# Schur decomposition
#import scipy
#print scipy.version.version
t,z=la.schur(a, output='real', lwork=None, overwrite_a=True, sort=None, check_finite=True)
print "schur form "
print t
print "orthogonal matrix "
print z


Yes you can do it via sticking a unitary transformation in the middle of the product hence we get
A = V * U * V^-1 = V * T' * T * U * T' * T * V^{-1}.
Once you get the idea you can optimize the code by tiling things but let's do it the naive way by forming T explicitly.
If the matrix is even-sized then all blocks are complex conjugates. Otherwise we get a zero as the eigenvalue. The eigenvalues are guaranteed to have zero real parts so the first thing is to clean up the noise and then order such that the zeros are on the upper left corner (arbitrary choice).
n = 5
a = np.random.rand(n,n)
a=a-np.transpose(a)
[u,v] = np.linalg.eig(a)
perm = np.argsort(np.abs(np.imag(u)))
unew = 1j*np.imag(u[perm])
Obviously, we need to reorder the eigenvector matrix too to keep things equivalent.
vnew = v[:,perm]
Now so far we did nothing other than reordering the middle eigenvalue matrix in the eigenvalue decomposition. Now we switch from complex form to real block diagonal form.
First we have to know how many zero eigenvalues there are
numblocks = np.flatnonzero(unew).size // 2
num_zeros = n - (2 * numblocks)
Then we basically, form another unitary transformation (complex this time) and stick it the same way
T = sp.linalg.block_diag(*[1.]*num_zeros,np.kron(1/np.sqrt(2)*np.eye(numblocks),np.array([[1.,1j],[1,-1j]])))
Eigs = np.real(T.conj().T.dot(np.diag(unew).dot(T)))
Evecs = np.real(vnew.dot(T))
This gives you the new real valued decomposition. So the code all in one place
n = 5
a = np.random.rand(n,n)
a=a-np.transpose(a)
[u,v] = np.linalg.eig(a)
perm = np.argsort(np.abs(np.imag(u)))
unew = 1j*np.imag(u[perm])
vnew = v[perm,:]
numblocks = np.flatnonzero(unew).size // 2
num_zeros = n - (2 * numblocks)
T = sp.linalg.block_diag(*[1.]*num_zeros,np.kron(1/np.sqrt(2)*np.eye(numblocks),np.array([[1.,1j],[1,-1j]])))
Eigs = np.real(T.conj().T.dot(np.diag(unew).dot(T)))
Evecs = np.real(vnew.dot(T))
print(np.allclose(Evecs.dot(Eigs.dot(np.linalg.inv(Evecs))) - a,np.zeros((n,n))))
gives True. Note that this is the naive way of obtaining the real spectral decomposition. There are lots of places where you need to keep track of numerical error accumulation.
Example output
Eigs
Out[379]:
array([[ 0. , 0. , 0. , 0. , 0. ],
[ 0. , 0. , -0.61882847, 0. , 0. ],
[ 0. , 0.61882847, 0. , 0. , 0. ],
[ 0. , 0. , 0. , 0. , -1.05097581],
[ 0. , 0. , 0. , 1.05097581, 0. ]])
Evecs
Out[380]:
array([[-0.15419078, -0.27710323, -0.39594838, 0.05427001, -0.51566173],
[-0.22985364, 0.0834649 , 0.23147553, -0.085043 , -0.74279915],
[ 0.63465436, 0.49265672, 0. , 0.20226271, -0.38686576],
[-0.02610706, 0.60684296, -0.17832525, 0.23822511, 0.18076858],
[-0.14115513, -0.23511356, 0.08856671, 0.94454277, 0. ]])

Related

Re-calculate elements of symmetric matrix using a "i not equal to j" loop in Python

The correlation matrix is a symmetric matrix, meaning that its upper diagonal and lower diagonal elements are mirror images of each other, together called off-diagonal elements (as opposed to the diagonal elements, which are all equal to 1 in any correlation matrix since any variable's correlation with itself is just 1).
The off-diagonal elements of a correlation matrix are the same wherever the i'th row number and j'th column number in the lower diagonal are swapped in the upper diagonal, i.e. correlation of variables 1 and 2 (row 1, column 2) are the same for variables 2 and 1 (row 2, column 1). Therefore, we only need to re-calculate the lower-diagonal elements, and copy them to corresponding positions in the matrix's upper-diagonal after
import numpy as np
from numpy.random import randn
X = randn(20,3)
Rho = np.corrcoef(X.T) #correlation matrix
print(np.tril(Rho)) #lower off-diagonal of matrix Rho to re-calculate, then copy to other side
shows
array([[ 1. , 0. , 0. ],
[-0.03003281, 1. , 0. ],
[-0.02602238, 0.06137713, 1. ]])
What is the most efficient way to code a "i not-equal-to j" loop for the following sequence of steps:
re-calculate the lower off-diagonal elements of the symmetric matrix according to some apply function (to make it simple, we will just add +2 to each of these elements)
flip those same calculations onto its mirror image (the corresponding upper off-diagonals)
Also, replace the diagonal elements of the symmetric matrix with a vector filled with 10's (instead of 1's as found in the correlation matrix)
The aim is to generate a new matrix that is a re-calculation of the original.

Let us generate Rho first (note that I'm initializing the pseudo-random number generator in order to obtain the same Rho in different runs of the code):
In [526]: import numpy as np
In [527]: np.random.seed(0)
...: n = 3
...: X = np.random.randn(20, n)
...: Rho = np.corrcoef(X.T)
In [528]: Rho
Out[528]:
array([[1. , 0.03224462, 0.05021998],
[0.03224462, 1. , 0.15140358],
[0.05021998, 0.15140358, 1. ]])
Then you can use NumPy's tril_indices_from and advanced indexing to generate the new matrix:
In [548]: result = np.zeros_like(Rho)
In [549]: lrows, lcols = np.tril_indices_from(Rho, k=-1)
In [550]: result[lrows, lcols] = Rho[lrows, lcols] + 2
In [551]: result
Out[551]:
array([[0. , 0. , 0. ],
[2.03224462, 0. , 0. ],
[2.05021998, 2.15140358, 0. ]])
In [552]: result[lcols, lrows] = result[lrows, lcols]
In [553]: result
Out[553]:
array([[0. , 2.03224462, 2.05021998],
[2.03224462, 0. , 2.15140358],
[2.05021998, 2.15140358, 0. ]])
In [554]: result[np.arange(n), np.arange(n)] = 10
In [555]: result
Out[555]:
array([[10. , 2.03224462, 2.05021998],
[ 2.03224462, 10. , 2.15140358],
[ 2.05021998, 2.15140358, 10. ]])

Quickest way to calculate the euclidean distance matrix of two list of points [duplicate]

I have a set of points in 2-dimensional space and need to calculate the distance from each point to each other point.
I have a relatively small number of points, maybe at most 100. But since I need to do it often and rapidly in order to determine the relationships between these moving points, and since I'm aware that iterating through the points could be as bad as O(n^2) complexity, I'm looking for ways to take advantage of numpy's matrix magic (or scipy).
As it stands in my code, the coordinates of each object are stored in its class. However, I could also update them in a numpy array when I update the class coordinate.
class Cell(object):
"""Represents one object in the field."""
def __init__(self,id,x=0,y=0):
self.m_id = id
self.m_x = x
self.m_y = y
It occurs to me to create a Euclidean distance matrix to prevent duplication, but perhaps you have a cleverer data structure.
I'm open to pointers to nifty algorithms as well.
Also, I note that there are similar questions dealing with Euclidean distance and numpy but didn't find any that directly address this question of efficiently populating a full distance matrix.

You can take advantage of the complex type :
# build a complex array of your cells
z = np.array([complex(c.m_x, c.m_y) for c in cells])
First solution
# mesh this array so that you will have all combinations
m, n = np.meshgrid(z, z)
# get the distance via the norm
out = abs(m-n)
Second solution
Meshing is the main idea. But numpy is clever, so you don't have to generate m & n. Just compute the difference using a transposed version of z. The mesh is done automatically :
out = abs(z[..., np.newaxis] - z)
Third solution
And if z is directly set as a 2-dimensional array, you can use z.T instead of the weird z[..., np.newaxis]. So finally, your code will look like this :
z = np.array([[complex(c.m_x, c.m_y) for c in cells]]) # notice the [[ ... ]]
out = abs(z.T-z)
Example
>>> z = np.array([[0.+0.j, 2.+1.j, -1.+4.j]])
>>> abs(z.T-z)
array([[ 0. , 2.23606798, 4.12310563],
[ 2.23606798, 0. , 4.24264069],
[ 4.12310563, 4.24264069, 0. ]])
As a complement, you may want to remove duplicates afterwards, taking the upper triangle :
>>> np.triu(out)
array([[ 0. , 2.23606798, 4.12310563],
[ 0. , 0. , 4.24264069],
[ 0. , 0. , 0. ]])
Some benchmarks
>>> timeit.timeit('abs(z.T-z)', setup='import numpy as np;z = np.array([[0.+0.j, 2.+1.j, -1.+4.j]])')
4.645645342274779
>>> timeit.timeit('abs(z[..., np.newaxis] - z)', setup='import numpy as np;z = np.array([0.+0.j, 2.+1.j, -1.+4.j])')
5.049334864854522
>>> timeit.timeit('m, n = np.meshgrid(z, z); abs(m-n)', setup='import numpy as np;z = np.array([0.+0.j, 2.+1.j, -1.+4.j])')
22.489568296184686

If you don't need the full distance matrix, you will be better off using kd-tree. Consider scipy.spatial.cKDTree or sklearn.neighbors.KDTree. This is because a kd-tree kan find k-nearnest neighbors in O(n log n) time, and therefore you avoid the O(n**2) complexity of computing all n by n distances.

Jake Vanderplas gives this example using broadcasting in Python Data Science Handbook, which is very similar to what #shx2 proposed.
import numpy as np
rand = random.RandomState(42)
X = rand.rand(3, 2)
dist_sq = np.sum((X[:, np.newaxis, :] - X[np.newaxis, :, :]) ** 2, axis = -1)
dist_sq
array([[0. , 0.18543317, 0.81602495],
[0.18543317, 0. , 0.22819282],
[0.81602495, 0.22819282, 0. ]])

Here is how you can do it using numpy:
import numpy as np
x = np.array([0,1,2])
y = np.array([2,4,6])
# take advantage of broadcasting, to make a 2dim array of diffs
dx = x[..., np.newaxis] - x[np.newaxis, ...]
dy = y[..., np.newaxis] - y[np.newaxis, ...]
dx
=> array([[ 0, -1, -2],
[ 1, 0, -1],
[ 2, 1, 0]])
# stack in one array, to speed up calculations
d = np.array([dx,dy])
d.shape
=> (2, 3, 3)
Now all is left is computing the L2-norm along the 0-axis (as discussed here):
(d**2).sum(axis=0)**0.5
=> array([[ 0. , 2.23606798, 4.47213595],
[ 2.23606798, 0. , 2.23606798],
[ 4.47213595, 2.23606798, 0. ]])

If you are looking for the most efficient way of computation - use SciPy's cdist() (or pdist() if you need just vector of pairwise distances instead of full distance matrix) as suggested in Tweakimp's comment. As he said it's a lot faster than method based on vectorization and broadcasting, proposed by RichPauloo and shx2. The reason for that is that SciPy's cdist() and pdist() under the hood use for loop and C implementations for metric computations, which are even faster than vectorization.
By the way, if you can use SciPy and still prefer method using broadcasting, you don't have to implement it by yourself, as distance_matrix() function is pure Python implementation, which leverages broadcasting and vectorization (source code, docs).
It's worth mentioning that cdist()/pdist() is also more efficient than broadcasting memory-wise, as it computes distances one by one and avoids creating arrays of n*n*d elements, where n is number of points and d is points' dimensionality.
Experiments
I've conducted some simple experiments to compare performance of SciPy's cdist(), distance_matrix() and broadcasting implementation in NumPy. I used perf_counter_ns() from Python's time module to measure time and all the results are averaged over 10 runs on 10000 points in 2D space using np.float64 datatype (tested on Python 3.8.10, Windows 10 with Ryzen 2700 and 16 GB RAM):
cdist() - 0.6724s
distance_matrix() - 3.0128s
my NumPy implementation - 3.6931s
Code if someone wants to reproduce experiments:
from scipy.spatial import *
import numpy as np
from time import perf_counter_ns
def dist_mat_custom(a, b):
return np.sqrt(np.sum(np.square(a[:, np.newaxis, :] - b[np.newaxis, :, :]), axis=-1))
results = []
size = 10000
it_num = 10
for i in range(it_num):
a = np.random.normal(size=(size, 2))
b = np.random.normal(size=(size, 2))
start = perf_counter_ns()
c = distance_matrix(a, b)
#c = dist_mat_custom(a, b)
#c = distance.cdist(a, b)
results.append(perf_counter_ns() - start)
print(np.mean(results) / 1e9)

Markov Clustering in Python

As the title says, I'm trying to get a Markov Clustering Algorithm to work in Python, namely Python 3.7
Unfortunately, it's not doing much of anything, and it's driving me up the wall trying to fix it.
EDIT: First, I've made the adjustments to the main code to make each column sum to 100, even if it's not perfectly balanced. I'm going to try to account for that in the final answer.
To be clear, the biggest problem is that the numbers spiral out of control, into such easily-understandable numbers as 5.56268465e-309, and I don't know how to convert that into something understandable.
Here's the code so far:
import numpy as np
import math
## How far you'd like your random-walkers to go (bigger number -> more walking)
EXPANSION_POWER = 2
## How tightly clustered you'd like your final picture to be (bigger number -> more clusters)
INFLATION_POWER = 2
ITERATION_COUNT = 10
def normalize(matrix):
return matrix/np.sum(matrix, axis=0)
def expand(matrix, power):
return np.linalg.matrix_power(matrix, power)
def inflate(matrix, power):
for entry in np.nditer(transition_matrix, op_flags=['readwrite']):
entry[...] = math.pow(entry, power)
return matrix
def run(matrix):
#np.fill_diagonal(matrix, 1)
#print(matrix)
matrix = normalize(matrix)
print(matrix)
for _ in range(ITERATION_COUNT):
matrix = normalize(inflate(expand(matrix, EXPANSION_POWER), INFLATION_POWER))
return matrix
transition_matrix = np.array ([[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0.5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0.5,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0.34,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0.33,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0.33,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0.34,0,0,0,0,0,0,0,0,0,0,0,0,0.125,0],
[0,0,0,0.33,0,0,0.5,0,0,0,0,0,0,0,0,0,0.125,1],
[0,0,0,0.33,0,0,0.5,1,1,0,0,0,0,0,0,0,0.125,0],
[0,0,0,0,0.166,0,0,0,0,0,0,0,0,0,0,0,0.125,0],
[0,0,0,0,0.166,0,0,0,0,0.2,0,0,0,0,0,0,0.125,0],
[0,0,0,0,0.167,0,0,0,0,0.2,0.25,0,0,0,0,0,0.125,0],
[0,0,0,0,0.167,0,0,0,0,0.2,0.25,0.5,0,0,0,0,0,0],
[0,0,0,0,0.167,0,0,0,0,0.2,0.25,0.5,0,1,0,0,0.125,0],
[0,0,0,0,0.167,0,0,0,0,0.2,0.25,0,1,0,1,0,0.125,0],
[0,0,0,0,0,0.34,0,0,0,0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0.33,0,0,0,0,0,0,0,0,0,0.5,0,0],
[0,0,0,0,0,0.33,0,0,0,0,0,0,0,0,0,0.5,0,0]])
run(transition_matrix)
print(transition_matrix)
This is part of a uni assignment - I need to do this array both weighted and unweighted (though the weighted part can just wait until I've got the bloody thing working at all) any tips or suggestions?

Your transition matrix is not valid.
>>> transition_matrix.sum(axis=0)
>>> matrix([[1. , 1. , 0.99, 0.99, 0.96, 0.99, 1. , 1. , 0. , 1. ,
1. , 1. , 1. , 0. , 0. , 1. , 0.88, 1. ]])
Not only does some of your columns not sum to 1, some of them sum to 0.
This means when you try to normalize your matrix, you will end up with nan because you are dividing by 0.
Lastly, is there a reason why you are using a Numpy matrix instead of just a Numpy array, which is the recommended container for such data? Because using Numpy arrays will simplify some of the operations, such as raising each entry to a power. Also, there are some differences between Numpy matrix and Numpy array which can result in subtle bugs.

How to blur 3D array of points, while maintaining their original values? (Python)

I have a sparse 3D array of values. I am trying to turn each "point" into a fuzzy "sphere", by applying a Gaussian filter to the array.
I would like the original value at the point (x,y,z) to remain the same. I just want to create falloff values around this point... But applying the Gaussian filter changes the original (x,y,z) value as well.
I am currently doing this:
dataCube = scipy.ndimage.filters.gaussian_filter(dataCube, 3, truncate=8)
Is there a way for me to normalize this, or do something so that my original values are still in this new dataCube? I am not necessarily tied to using a Gaussian filter, if that is not the best approach.

You can do this using a convolution with a kernel that has 1 as its central value, and a width smaller than the spacing between your data points.
1-d example:
import numpy as np
import scipy.signal
data = np.array([0,0,0,0,0,5,0,0,0,0,0])
kernel = np.array([0.5,1,0.5])
scipy.signal.convolve(data, kernel, mode="same")
gives
array([ 0. , 0. , 0. , 0. , 2.5, 5. , 2.5, 0. , 0. , 0. , 0. ])
Note that fftconvolve might be much faster for large arrays. You also have to specify what should happen at the boundaries of your array.
Update: 3-d example
import numpy as np
from scipy import signal
# first build the smoothing kernel
sigma = 1.0 # width of kernel
x = np.arange(-3,4,1) # coordinate arrays -- make sure they contain 0!
y = np.arange(-3,4,1)
z = np.arange(-3,4,1)
xx, yy, zz = np.meshgrid(x,y,z)
kernel = np.exp(-(xx**2 + yy**2 + zz**2)/(2*sigma**2))
# apply to sample data
data = np.zeros((11,11,11))
data[5,5,5] = 5.
filtered = signal.convolve(data, kernel, mode="same")
# check output
print filtered[:,5,5]
gives
[ 0. 0. 0.05554498 0.67667642 3.0326533 5. 3.0326533
0.67667642 0.05554498 0. 0. ]

Efficiently Calculating a Euclidean Distance Matrix Using Numpy

I have a set of points in 2-dimensional space and need to calculate the distance from each point to each other point.
I have a relatively small number of points, maybe at most 100. But since I need to do it often and rapidly in order to determine the relationships between these moving points, and since I'm aware that iterating through the points could be as bad as O(n^2) complexity, I'm looking for ways to take advantage of numpy's matrix magic (or scipy).
As it stands in my code, the coordinates of each object are stored in its class. However, I could also update them in a numpy array when I update the class coordinate.
class Cell(object):
"""Represents one object in the field."""
def __init__(self,id,x=0,y=0):
self.m_id = id
self.m_x = x
self.m_y = y
It occurs to me to create a Euclidean distance matrix to prevent duplication, but perhaps you have a cleverer data structure.
I'm open to pointers to nifty algorithms as well.
Also, I note that there are similar questions dealing with Euclidean distance and numpy but didn't find any that directly address this question of efficiently populating a full distance matrix.

You can take advantage of the complex type :
# build a complex array of your cells
z = np.array([complex(c.m_x, c.m_y) for c in cells])
First solution
# mesh this array so that you will have all combinations
m, n = np.meshgrid(z, z)
# get the distance via the norm
out = abs(m-n)
Second solution
Meshing is the main idea. But numpy is clever, so you don't have to generate m & n. Just compute the difference using a transposed version of z. The mesh is done automatically :
out = abs(z[..., np.newaxis] - z)
Third solution
And if z is directly set as a 2-dimensional array, you can use z.T instead of the weird z[..., np.newaxis]. So finally, your code will look like this :
z = np.array([[complex(c.m_x, c.m_y) for c in cells]]) # notice the [[ ... ]]
out = abs(z.T-z)
Example
>>> z = np.array([[0.+0.j, 2.+1.j, -1.+4.j]])
>>> abs(z.T-z)
array([[ 0. , 2.23606798, 4.12310563],
[ 2.23606798, 0. , 4.24264069],
[ 4.12310563, 4.24264069, 0. ]])
As a complement, you may want to remove duplicates afterwards, taking the upper triangle :
>>> np.triu(out)
array([[ 0. , 2.23606798, 4.12310563],
[ 0. , 0. , 4.24264069],
[ 0. , 0. , 0. ]])
Some benchmarks
>>> timeit.timeit('abs(z.T-z)', setup='import numpy as np;z = np.array([[0.+0.j, 2.+1.j, -1.+4.j]])')
4.645645342274779
>>> timeit.timeit('abs(z[..., np.newaxis] - z)', setup='import numpy as np;z = np.array([0.+0.j, 2.+1.j, -1.+4.j])')
5.049334864854522
>>> timeit.timeit('m, n = np.meshgrid(z, z); abs(m-n)', setup='import numpy as np;z = np.array([0.+0.j, 2.+1.j, -1.+4.j])')
22.489568296184686

If you don't need the full distance matrix, you will be better off using kd-tree. Consider scipy.spatial.cKDTree or sklearn.neighbors.KDTree. This is because a kd-tree kan find k-nearnest neighbors in O(n log n) time, and therefore you avoid the O(n**2) complexity of computing all n by n distances.

Jake Vanderplas gives this example using broadcasting in Python Data Science Handbook, which is very similar to what #shx2 proposed.
import numpy as np
rand = random.RandomState(42)
X = rand.rand(3, 2)
dist_sq = np.sum((X[:, np.newaxis, :] - X[np.newaxis, :, :]) ** 2, axis = -1)
dist_sq
array([[0. , 0.18543317, 0.81602495],
[0.18543317, 0. , 0.22819282],
[0.81602495, 0.22819282, 0. ]])

Here is how you can do it using numpy:
import numpy as np
x = np.array([0,1,2])
y = np.array([2,4,6])
# take advantage of broadcasting, to make a 2dim array of diffs
dx = x[..., np.newaxis] - x[np.newaxis, ...]
dy = y[..., np.newaxis] - y[np.newaxis, ...]
dx
=> array([[ 0, -1, -2],
[ 1, 0, -1],
[ 2, 1, 0]])
# stack in one array, to speed up calculations
d = np.array([dx,dy])
d.shape
=> (2, 3, 3)
Now all is left is computing the L2-norm along the 0-axis (as discussed here):
(d**2).sum(axis=0)**0.5
=> array([[ 0. , 2.23606798, 4.47213595],
[ 2.23606798, 0. , 2.23606798],
[ 4.47213595, 2.23606798, 0. ]])

If you are looking for the most efficient way of computation - use SciPy's cdist() (or pdist() if you need just vector of pairwise distances instead of full distance matrix) as suggested in Tweakimp's comment. As he said it's a lot faster than method based on vectorization and broadcasting, proposed by RichPauloo and shx2. The reason for that is that SciPy's cdist() and pdist() under the hood use for loop and C implementations for metric computations, which are even faster than vectorization.
By the way, if you can use SciPy and still prefer method using broadcasting, you don't have to implement it by yourself, as distance_matrix() function is pure Python implementation, which leverages broadcasting and vectorization (source code, docs).
It's worth mentioning that cdist()/pdist() is also more efficient than broadcasting memory-wise, as it computes distances one by one and avoids creating arrays of n*n*d elements, where n is number of points and d is points' dimensionality.
Experiments
I've conducted some simple experiments to compare performance of SciPy's cdist(), distance_matrix() and broadcasting implementation in NumPy. I used perf_counter_ns() from Python's time module to measure time and all the results are averaged over 10 runs on 10000 points in 2D space using np.float64 datatype (tested on Python 3.8.10, Windows 10 with Ryzen 2700 and 16 GB RAM):
cdist() - 0.6724s
distance_matrix() - 3.0128s
my NumPy implementation - 3.6931s
Code if someone wants to reproduce experiments:
from scipy.spatial import *
import numpy as np
from time import perf_counter_ns
def dist_mat_custom(a, b):
return np.sqrt(np.sum(np.square(a[:, np.newaxis, :] - b[np.newaxis, :, :]), axis=-1))
results = []
size = 10000
it_num = 10
for i in range(it_num):
a = np.random.normal(size=(size, 2))
b = np.random.normal(size=(size, 2))
start = perf_counter_ns()
c = distance_matrix(a, b)
#c = dist_mat_custom(a, b)
#c = distance.cdist(a, b)
results.append(perf_counter_ns() - start)
print(np.mean(results) / 1e9)

Categories