I have my X and Y numpy arrays:
X = np.array([0,1,2,3])
Y = np.array([0,1,2,3])
And my function which maps x,y values to Z points:
def z(x,y):
return x+y
I wish to produce the obvious thing required for a 3D plot: the 2-dimensional numpy array for the corresponding Z-values. I believe it should look like:
Z = np.array([[0, 1, 2, 3],
[1, 2, 3, 4],
[2, 3, 4, 5],
[3, 4, 5, 6]])
I can do this in several lines, but I'm looking for the briefest most elegant piece of code.
For a function that is array aware it is more economical to use an open grid:
>>> import numpy as np
>>>
>>> X = np.array([0,1,2,3])
>>> Y = np.array([0,1,2,3])
>>>
>>> def z(x,y):
... return x+y
...
>>> XX, YY = np.ix_(X, Y)
>>> XX, YY
(array([[0],
[1],
[2],
[3]]), array([[0, 1, 2, 3]]))
>>> z(XX, YY)
array([[0, 1, 2, 3],
[1, 2, 3, 4],
[2, 3, 4, 5],
[3, 4, 5, 6]])
If your grid axes are ranges you can directly create the grid using np.ogrid
>>> XX, YY = np.ogrid[:4, :4]
>>> XX, YY
(array([[0],
[1],
[2],
[3]]), array([[0, 1, 2, 3]]))
If the function is not array aware you can make it so using np.vectorize:
>>> def f(x, y):
... if x > y:
... return x
... else:
... return -x
...
>>> np.vectorize(f)(*np.ogrid[-3:4, -3:4])
array([[ 3, 3, 3, 3, 3, 3, 3],
[-2, 2, 2, 2, 2, 2, 2],
[-1, -1, 1, 1, 1, 1, 1],
[ 0, 0, 0, 0, 0, 0, 0],
[ 1, 1, 1, 1, -1, -1, -1],
[ 2, 2, 2, 2, 2, -2, -2],
[ 3, 3, 3, 3, 3, 3, -3]])
One very short way to achieve what you want is to produce a meshgrid from your coordinates:
X,Y = np.meshgrid(x,y)
z = X+Y
or more general:
z = f(X,Y)
or even in one line:
z = f(*np.meshgrid(x,y))
EDIT:
If your function also may return a constant, you have to somehow infer the dimensions that the result should have. If you want to continue using meshgrids one very simple way would be re-write your function in this way:
def f(x,y):
return x*0+y*0+a
where a would be your constant. numpy would then take care of the dimensions for you. This is of course a bit weird looking, so instead you could write
def f(x,y):
return np.full(x.shape, a)
If you really want to go with functions that work both on scalars and arrays, it's probably best to go with np.vectorize as in #PaulPanzer's answer.
Related
Given two arrays, a and b, how to find efficiently all combinations of elements in b that have equal value in a?
here is an example:
Given
a = [0, 0, 0, 1, 1, 2, 2, 2, 2]
b = [1, 2, 4, 5, 9, 3, 7, 22, 10]
how would you calculate
c = [[1, 2],
[1, 4],
[2, 4],
[5, 9],
[3, 7],
[3, 22],
[3, 10],
[7, 22],
[7, 10],
[22, 10]]
?
a can be assumed to be sorted.
I can do this with loops, a la:
import torch
a = torch.tensor([0, 0, 0, 1, 1, 2, 2, 2, 2])
b = torch.tensor([1, 2, 4, 5, 9, 3, 7, 22, 10])
jumps = torch.cat((torch.tensor([0]),
torch.where(a.diff() > 0)[0] + 1,
torch.tensor([len(a)])))
cs = []
for i in range(len(jumps) - 1):
cs.append(torch.combinations(b[jumps[i]:jumps[i + 1]]))
c = torch.cat(cs)
Is there any efficient way to avoid the loop? The solution should work for CPU and CUDA.
Also, the solution should have runtime O(m * m), where m is the largest number of equal elements in a and not O(n * n) where n is the length of of a.
I prefer solutions for pytorch, but I am curious for solution for numpy as well.
I think the overhead of using torch is only justified for bigger datasets, as there is basically no computational difficulty in the function, imho you can achieve same results with:
from collections import Counter
def find_combinations1(a, b):
count_a = Counter(a)
combinations = []
for x in set(b):
if count_a[x] == b.count(x):
combinations.append(x)
return combinations
or even a simpler:
def find_combinations2(a, b):
return list(set(a) & set(b))
With pytorch I assume the most simple approach is:
import torch
def find_combinations3(a, b):
a = torch.tensor(a)
b = torch.tensor(b)
eq = torch.eq(a, b.view(-1, 1))
indices = torch.nonzero(eq)
return indices[:, 1]
This option has of course a time complexity of O(n*m) where n is the size of a and m is the size of b, and O(n+m) is the memory for the tensors.
I would like to add the values in the above row to the row below using vectorization. For example, if I had the ndarray,
[[0, 0, 0, 0],
[1, 1, 1, 1],
[2, 2, 2, 2],
[3, 3, 3, 3]]
Then after one iteration through this method, it would result in
[[0, 0, 0, 0],
[1, 1, 1, 1],
[3, 3, 3, 3],
[5, 5, 5, 5]]
One can simply do this with a for loop:
import numpy as np
def addAboveRow(arr):
cpy = arr.copy()
r, c = arr.shape
for i in range(1, r):
for j in range(c):
cpy[i][j] += arr[i - 1][j]
return cpy
ndarr = np.array([0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3]).reshape(4, 4)
print(addAboveRow(ndarr))
I'm not sure how to approach this using vectorization though. I think slicers should be used? Also, I'm not really sure how to deal with the issue of the top border, because nothing should be added onto the first row. Any help would be appreciated. Thanks!
Note: I am really new to vectorization so an explanation would be great!
You can use indexing directly:
b = np.zeros_like(a)
b[0] = a[0]
b[1:] = a[1:] + a[:-1]
>>> b
array([[0, 0, 0, 0],
[1, 1, 1, 1],
[3, 3, 3, 3],
[5, 5, 5, 5]])
An alternative:
b = a.copy()
b[1:] += a[:-1]
Or:
b = a.copy()
np.add(b[1:], a[:-1], out=b[1:])
You could try the following
np.put(arr, np.arange(arr.shape[1], arr.size), arr[1:]+arr[:-1])
I have an array such as the following:
In [70]: x
Out[70]:
array([[0, 1, 2],
[3, 4, 5]])
I am trying to get the indices per row where a condition holds, for example, x > 1.
Expected output is like ([2], [0, 1, 2])
I have tried numpy.where, numpy.nonzero, but they give strange results.
One approach -
r,c = np.where(x>1)
out = np.split(c, np.flatnonzero(r[1:] > r[:-1])+1)
Sample run -
In [140]: x
Out[140]:
array([[0, 2, 0, 1, 1],
[2, 2, 1, 2, 0],
[0, 2, 1, 1, 0],
[1, 0, 0, 2, 2]])
In [141]: r,c = np.where(x>1)
In [142]: np.split(c, np.flatnonzero(r[1:] > r[:-1])+1)
Out[142]: [array([1]), array([0, 1, 3]), array([1]), array([3, 4])]
Alternatively, we could use np.unique on the final step, like so -
np.split(c, np.unique(r, return_index=1)[1][1:])
I have a 3d array
A = np.random.random((4,4,3))
and a index matrix
B = np.int_(np.random.random((4,4))*3)
How do I get a 2D array from A based on index matrix B?
In general, how to get a N-1 dimensional array from a ND array and a N-1 dimensional index array?
Lets take an example:
>>> A = np.random.randint(0,10,(3,3,2))
>>> A
array([[[0, 1],
[8, 2],
[6, 4]],
[[1, 0],
[6, 9],
[7, 7]],
[[1, 2],
[2, 2],
[9, 7]]])
Use fancy indexing to take simple indices. Note that the all indices must be of the same shape and the shape of each index will be what is returned.
>>> ind = np.arange(2)
>>> A[ind,ind,ind]
array([0, 9]) #Index (0,0,0) and (1,1,1)
>>> ind = np.arange(2).reshape(2,1)
>>> A[ind,ind,ind]
array([[0],
[9]])
So for your example we need to supply the grid for the first two dimensions:
>>> A = np.random.random((4,4,3))
>>> B = np.int_(np.random.random((4,4))*3)
>>> A
array([[[ 0.95158697, 0.37643036, 0.29175815],
[ 0.84093397, 0.53453123, 0.64183715],
[ 0.31189496, 0.06281937, 0.10008886],
[ 0.79784114, 0.26428462, 0.87899921]],
[[ 0.04498205, 0.63823379, 0.48130828],
[ 0.93302194, 0.91964805, 0.05975115],
[ 0.55686047, 0.02692168, 0.31065731],
[ 0.92822499, 0.74771321, 0.03055592]],
[[ 0.24849139, 0.42819062, 0.14640117],
[ 0.92420031, 0.87483486, 0.51313695],
[ 0.68414428, 0.86867423, 0.96176415],
[ 0.98072548, 0.16939697, 0.19117458]],
[[ 0.71009607, 0.23057644, 0.80725518],
[ 0.01932983, 0.36680718, 0.46692839],
[ 0.51729835, 0.16073775, 0.77768313],
[ 0.8591955 , 0.81561797, 0.90633695]]])
>>> B
array([[1, 2, 0, 0],
[1, 2, 0, 1],
[2, 1, 1, 1],
[1, 2, 1, 2]])
>>> x,y = np.meshgrid(np.arange(A.shape[0]),np.arange(A.shape[1]))
>>> x
array([[0, 1, 2, 3],
[0, 1, 2, 3],
[0, 1, 2, 3],
[0, 1, 2, 3]])
>>> y
array([[0, 0, 0, 0],
[1, 1, 1, 1],
[2, 2, 2, 2],
[3, 3, 3, 3]])
>>> A[x,y,B]
array([[ 0.37643036, 0.48130828, 0.24849139, 0.71009607],
[ 0.53453123, 0.05975115, 0.92420031, 0.36680718],
[ 0.10008886, 0.02692168, 0.86867423, 0.16073775],
[ 0.26428462, 0.03055592, 0.16939697, 0.90633695]])
If you prefer to use mesh as suggested by Daniel, you may also use
A[tuple( np.ogrid[:A.shape[0], :A.shape[1]] + [B] )]
to work with sparse indices. In the general case you could use
A[tuple( np.ogrid[ [slice(0, end) for end in A.shape[:-1]] ] + [B] )]
Note that this may also be used when you'd like to index by B an axis different from the last one (see for example this answer about inserting an element into a list).
Otherwise you can do it using broadcasting:
A[np.arange(A.shape[0])[:, np.newaxis], np.arange(A.shape[1])[np.newaxis, :], B]
This may be generalized too but it's a bit more complicated.
You have an original sparse matrix X:
>>print type(X)
>>print X.todense()
<class 'scipy.sparse.csr.csr_matrix'>
[[1,4,3]
[3,4,1]
[2,1,1]
[3,6,3]]
You have a second sparse matrix Z, which is derived from some rows of X (say the values are doubled so we can see the difference between the two matrices). In pseudo-code:
>>Z = X[[0,2,3]]
>>print Z.todense()
[[1,4,3]
[2,1,1]
[3,6,3]]
>>Z = Z*2
>>print Z.todense()
[[2, 8, 6]
[4, 2, 2]
[6, 12,6]]
What's the best way of retrieving the rows in Z using the ORIGINAL indices from X. So for instance, in pseudo-code:
>>print Z[[0,3]]
[[2,8,6] #0 from Z, and what would be row **0** from X)
[6,12,6]] #2 from Z, but what would be row **3** from X)
That is, how can you retrieve rows from Z, using indices that refer to the original rows position in the original matrix X? To do this, you can't modify X in anyway (you can't add an index column to the matrix X), but there are no other limits.
If you have the original indices in an array i, and the values in i are in increasing order (as in your example), you can use numpy.searchsorted(i, [0, 3]) to find the indices in Z that correspond to indices [0, 3] in the original X. Here's a demonstration in an IPython session:
In [39]: X = csr_matrix([[1,4,3],[3,4,1],[2,1,1],[3,6,3]])
In [40]: X.todense()
Out[40]:
matrix([[1, 4, 3],
[3, 4, 1],
[2, 1, 1],
[3, 6, 3]])
In [41]: i = array([0, 2, 3])
In [42]: Z = 2 * X[i]
In [43]: Z.todense()
Out[43]:
matrix([[ 2, 8, 6],
[ 4, 2, 2],
[ 6, 12, 6]])
In [44]: Zsub = Z[searchsorted(i, [0, 3])]
In [45]: Zsub.todense()
Out[45]:
matrix([[ 2, 8, 6],
[ 6, 12, 6]])