We Keep Coding

Vectorization: Each row of the mask contains the column indices to mask for the corresponding row of the array

I have an array and a mask array. They have the same rows. Each row of the mask contains the indices to mask the array for the corresponding row. How to do the vectorization instead of using for loop?
Codes like this:
a = np.zeros((2, 4))
mask = np.array([[2, 3], [0, 1]])
# I'd like a vectorized way to do this (because the rows and cols are large):
a[0, mask[0]] = 1
a[1, mask[1]] = 1
This is what I want to obtain:
array([[0., 0., 1., 1.],
[1., 1., 0., 0.]])
==================================
The question has been answered by #mozway, but the efficiency between the for-loop solution and vectorized one is questioned by #AhmedAEK. So I did the efficiency comparison:
N = 5000
M = 10000
a = np.zeros((N, M))
# choice without replacement
mask = np.random.rand(N, M).argpartition(3, axis=1)[:,:3]
def t1():
for i in range(N):
a[i, mask[i]] = 1
def t2():
a[np.arange(a.shape[0])[:, None], mask] = 1
Then I use %timeit in Jupyter and got this screenshot:

You can use:
a[[[0],[1]], mask] = 1
Or, programmatically generating the rows slicer:
a[np.arange(a.shape[0])[:,None], mask] = 1
output:
array([[0., 0., 1., 1.],
[1., 1., 0., 0.]])

Select array elements with variable index bounds in numpy

This might be not possible as the intermediate array would have variable length rows.
What I am trying to accomplish is assigning a value to an array for the elements which have ad index delimited by my array of bounds. As an example:
bounds = np.array([[1,2], [1,3], [1,4]])
array = np.zeros((3,4))
__assign(array, bounds, 1)
after the assignment should result in
array = [
[0, 1, 0, 0],
[0, 1, 1, 0],
[0, 1, 1, 1]
]
I have tried something like this in various iterations without success:
ind = np.arange(array.shape[0])
array[ind, bounds[ind][0]:bounds[ind][1]] = 1
I am trying to avoid loops as this function will be called a lot. Any ideas?

I'm by no means a Numpy expert, but from the different array indexing options I could find, this was the fastest solution I could figure out:
bounds = np.array([[1,2], [1,3], [1,4]])
array = np.zeros((3,4))
for i, x in enumerate(bounds):
cols = slice(x[0], x[1])
array[i, cols] = 1
Here we iterate through the list of bounds and reference the columns using slices.
I tried the below way of first constructing a list of column indices and a list of row indices, but it was way slower. Like 10 seconds plus vir 0.04 seconds on my laptop for a 10 000 x 10 000 array. I guess the slices make a huge difference.
bounds = np.array([[1,2], [1,3], [1,4]])
array = np.zeros((3,4))
cols = []
rows = []
for i, x in enumerate(bounds):
cols += list(range(x[0], x[1]))
rows += (x[1] - x[0]) * [i]
# print(cols) [1, 1, 2, 1, 2, 3]
# print(rows) [0, 1, 1, 2, 2, 2]
array[rows, cols] = 1

One of the issues with a purely NumPy method to solve this is that there exists no method to 'slice' a NumPy array using bounds from another over an axis. So the resultant expanded bounds end up becoming a variable-length list of lists such as [[1],[1,2],[1,2,3]. Then you can use np.eye and np.sum over axis=0 to get the required output.
bounds = np.array([[1,2], [1,3], [1,4]])
result = np.stack([np.sum(np.eye(4)[slice(*i)], axis=0) for i in bounds])
print(result)
array([[0., 1., 0., 0.],
[0., 1., 1., 0.],
[0., 1., 1., 1.]])
I tried various ways of being able to slice the np.eye(4) from [start:stop] over a NumPy array of starts and stops but sadly you will need an iteration to accomplish this.
EDIT: Another way you can do this in a vectorized way without any loops is -
def f(b):
o = np.sum(np.eye(4)[b[0]:b[1]], axis=0)
return o
np.apply_along_axis(f, 1, bounds)
array([[0., 1., 0., 0.],
[0., 1., 1., 0.],
[0., 1., 1., 1.]])
EDIT: If you are looking for a superfast solution but can tolerate a single for loop then the fastest approach based on my simulations among all answers on this thread is -
def h(bounds):
zz = np.zeros((len(bounds), bounds.max()))
for z,b in zip(zz,bounds):
z[b[0]:b[1]]=1
return zz
h(bounds)
array([[0., 1., 0., 0.],
[0., 1., 1., 0.],
[0., 1., 1., 1.]])

Using numba.njit decorator
import numpy as np
import numba
#numba.njit
def numba_assign_in_range(arr, bounds, val):
for i in range(len(bounds)):
s, e = bounds[i]
arr[i, s:e] = val
return arr
test_size = int(1e6) * 2
bounds = np.zeros((test_size, 2), dtype='int32')
bounds[:, 0] = 1
bounds[:, 1] = np.random.randint(0, 100, test_size)
a = np.zeros((test_size, 100))
with numba.njit
CPU times: user 3 µs, sys: 1 µs, total: 4 µs
Wall time: 6.2 µs
without numba.njit
CPU times: user 3.54 s, sys: 1.63 ms, total: 3.54 s
Wall time: 3.55 s

How to assign values to given indices to an array and average on repeated indices?

Is there a neat way to assign values to given indices in an array, and average values in repeated indices?
For example:
a = np.array([0, 0, 0, 0, 0])
ind = np.array([1, 1, 2, 3])
b = np.array([2, 3, 4, 5])
and I want to assign values in array b to array a at corresponding indices indicated in 'ind', and a[1] should be average of 2 and 3.
I can try a for-loop:
hit = np.zeros_like(a)
for i in range(ind.size):
hit[ind[i]] += 1
a[ind[i]] += b[i]
a = a / hit
But this code looks dirty. Is there any better way to do the job?

You could do this using np.where.
import numpy as np
a = np.array([0, 0, 0, 0, 0]).astype('float64')
ind = np.array([1, 1, 2, 3])
b = np.array([2, 3, 4, 5])
for i in set(ind):
a[i] = np.mean(b[np.where(ind == i)])
Would result in:
In [5]: a
Out[5]: array([0. , 2.5, 4. , 5. , 0. ])
You are essentially finding all indices of ind where the value of ind[index] is equal to i and then obtaining the mean of the values at those indices in b and assigning that mean to a[i]. Hope this helps!

Here is a vectorized method. The actual logic is close to your own solution.
n,d = (np.bincount(ind,x,a.size) for x in (b,None))
valid = d!=0
np.copyto(a,np.divide(n,d,where=valid),where=valid)

In [56]: a = np.zeros(5)
...: hit = np.zeros_like(a)
...: for i in range(ind.size):
...: hit[ind[i]] += 1
...: a[ind[i]] += b[i]
In [57]: a
Out[57]: array([0., 5., 4., 5., 0.])
In [58]: hit
Out[58]: array([0., 2., 1., 1., 0.])
The mention of duplicate indices brings to mind the .at ufunc method:
In [59]: a = np.zeros(5)
In [60]: a = np.zeros(5)
...: hit = np.zeros_like(a)
...: np.add.at(a,ind,b)
...: np.add.at(hit,ind,1)
In [61]: a
Out[61]: array([0., 5., 4., 5., 0.])
In [62]: hit
Out[62]: array([0., 2., 1., 1., 0.])
This isn't quite as fast as a[ind]=b, but faster than your loop.
np.bincount might well be better for this task, but this add.at is worth knowing and testing.
https://docs.scipy.org/doc/numpy/reference/generated/numpy.ufunc.at.html

This might not necessarily be cleaner or faster, but here's an alternative that I think is easy to read:
a = [[] for _ in range(5)]
for i, x in zip(ind, b):
a[i].append(x)
[np.mean(x) if len(x) else 0 for x in a]

Numpy indexing set 1 to max value and zero's to all others

I think I've misunderstood something with indexing in numpy.
I have a 3D-numpy array of shape (dim_x, dim_y, dim_z) and I want to find the maximum along the third axis (dim_z), and set its value to 1 and all the others to zero.
The problem is that I end up with several 1 in the same row, even if values are different.
Here is the code :
>>> test = np.random.rand(2,3,2)
>>> test
array([[[ 0.13110146, 0.07138861],
[ 0.84444158, 0.35296986],
[ 0.97414498, 0.63728852]],
[[ 0.61301975, 0.02313646],
[ 0.14251848, 0.91090492],
[ 0.14217992, 0.41549218]]])
>>> result = np.zeros_like(test)
>>> result[:test.shape[0], np.arange(test.shape[1]), np.argmax(test, axis=2)]=1
>>> result
array([[[ 1., 0.],
[ 1., 1.],
[ 1., 1.]],
[[ 1., 0.],
[ 1., 1.],
[ 1., 1.]]])
I was expecting to end with :
array([[[ 1., 0.],
[ 1., 0.],
[ 1., 0.]],
[[ 1., 0.],
[ 0., 1.],
[ 0., 1.]]])
Probably I'm missing something here. From what I've understood, 0:dim_x, np.arange(dim_y) returns dim_x of dim_y tuples and np.argmax(test, axis=dim_z) has the shape (dim_x, dim_y) so if the indexing is of the form [x, y, z] a couple [x, y] is not supposed to appear twice.
Could someone explain me where I'm wrong ? Thanks in advance.

What we are looking for
We get the argmax indices along the last axis -
idx = np.argmax(test, axis=2)
For the given sample data, we have idx :
array([[0, 0, 0],
[0, 1, 1]])
Now, idx covers the first and second axes, while getting those argmax indices.
To assign the corresponding ones in the output, we need to create range arrays for the first two axes covering the lengths along those and aligned according to the shape of idx. Now, idx is a 2D array of shape (m,n), where m = test.shape[0] and n = test.shape[1].
Thus, the range arrays for assignment into first two axes of output must be -
X = np.arange(test.shape[0])[:,None]
Y = np.arange(test.shape[1])
Notice, the extension of the first range array to 2D is needed to have it aligned against the rows of idx and Y would align against the cols of idx -
In [239]: X
Out[239]:
array([[0],
[1]])
In [240]: Y
Out[240]: array([0, 1, 2])
Schematically put -
idx :
Y array
--------->
x x x | X array
x x x |
v
The fault in original code
Your code was -
result[:test.shape[0], np.arange(test.shape[1]), ..
This is essentially :
result[:, np.arange(test.shape[1]), ...
So, you are selecting all elements along the first axis, instead of only selecting the corresponding ones that correspond to idx indices. In that process, you were selecting a lot more than required elements for assignment and hence you were seeing many more than required 1s in result array.
The correction
Thus, the only correction needed was indexing into the first axis with the range array and a working solution would be -
result[np.arange(test.shape[0])[:,None], np.arange(test.shape[1]), ...
The alternative(s)
Alternatively, using the range arrays created earlier with X and Y -
result[X,Y,idx] = 1
Another way to get X,Y would be with np.mgrid -
m,n = test.shape[:2]
X,Y = np.ogrid[:m,:n]

I think there's a problem with mixing basic (slice) and advanced indexing. It's easier to see when selecting value from an array than with this assignment; but it can result in transposed axes. For a problem like this it is better use advanced indexing all around, as provided by ix_
In [24]: test = np.random.rand(2,3,2)
In [25]: idx=np.argmax(test,axis=2)
In [26]: idx
Out[26]:
array([[1, 0, 1],
[0, 1, 1]], dtype=int32)
with basic and advanced:
In [31]: res1 = np.zeros_like(test)
In [32]: res1[:, np.arange(test.shape[1]), idx]=1
In [33]: res1
Out[33]:
array([[[ 1., 1.],
[ 1., 1.],
[ 0., 1.]],
[[ 1., 1.],
[ 1., 1.],
[ 0., 1.]]])
with advanced:
In [35]: I,J = np.ix_(range(test.shape[0]), range(test.shape[1]))
In [36]: I
Out[36]:
array([[0],
[1]])
In [37]: J
Out[37]: array([[0, 1, 2]])
In [38]: res2 = np.zeros_like(test)
In [40]: res2[I, J , idx]=1
In [41]: res2
Out[41]:
array([[[ 0., 1.],
[ 1., 0.],
[ 0., 1.]],
[[ 1., 0.],
[ 0., 1.],
[ 0., 1.]]])
On further thought, the use of the slice for the 1st dimension is just wrong , if the goal is to set or find the 6 argmax values
In [54]: test
Out[54]:
array([[[ 0.15288242, 0.36013289],
[ 0.90794601, 0.15265616],
[ 0.34014976, 0.53804266]],
[[ 0.97979479, 0.15898605],
[ 0.04933804, 0.89804999],
[ 0.10199319, 0.76170911]]])
In [55]: test[I, J, idx]
Out[55]:
array([[ 0.36013289, 0.90794601, 0.53804266],
[ 0.97979479, 0.89804999, 0.76170911]])
In [56]: test[:, J, idx]
Out[56]:
array([[[ 0.36013289, 0.90794601, 0.53804266],
[ 0.15288242, 0.15265616, 0.53804266]],
[[ 0.15898605, 0.04933804, 0.76170911],
[ 0.97979479, 0.89804999, 0.76170911]]])
With the slice it selects a (2,3,2) set of values from test (or res), not the intended (2,3). There 2 extra rows.

Here is an easier way to do it:
>>> test == test.max(axis=2, keepdims=1)
array([[[ True, False],
[ True, False],
[ True, False]],
[[ True, False],
[False, True],
[False, True]]], dtype=bool)
...and if you really want that as floating-point 1.0 and 0.0, then convert it:
>>> (test==test.max(axis=2, keepdims=1)).astype(float)
array([[[ 1., 0.],
[ 1., 0.],
[ 1., 0.]],
[[ 1., 0.],
[ 0., 1.],
[ 0., 1.]]])
Here is a way to do it with only one winner per row-column combo (i.e. no ties, as discussed in comments):
rowmesh, colmesh = np.meshgrid(range(test.shape[0]), range(test.shape[1]), indexing='ij')
maxloc = np.argmax(test, axis=2)
flatind = np.ravel_multi_index( [rowmesh, colmesh, maxloc ], test.shape )
result = np.zeros_like(test)
result.flat[flatind] = 1
UPDATE after reading hpaulj's answer:
rowmesh, colmesh = np.ix_(range(test.shape[0]), range(test.shape[1]))
is a more-efficient, more numpythonic, alternative to my meshgrid call (the rest of the code stays the same)
The issue of why your approach fails is hard to explain, but here's one place where intuition could start: your slicing approach says "all rows, times all columns, times a certain sequence of layers". How many elements is that slice in total? By contrast, how many elements do you actually want to set to 1? It can be instructive to look at the values you get when you view the corresponding test values of the slice you're trying to assign to:
>>> test[:, :, maxloc].shape
(2, 3, 2, 3) # oops! it's because maxloc itself is 2x3
>>> test[:, :, maxloc]
array([[[[ 0.13110146, 0.13110146, 0.13110146],
[ 0.13110146, 0.07138861, 0.07138861]],
[[ 0.84444158, 0.84444158, 0.84444158],
[ 0.84444158, 0.35296986, 0.35296986]],
[[ 0.97414498, 0.97414498, 0.97414498],
[ 0.97414498, 0.63728852, 0.63728852]]],
[[[ 0.61301975, 0.61301975, 0.61301975],
[ 0.61301975, 0.02313646, 0.02313646]],
[[ 0.14251848, 0.14251848, 0.14251848],
[ 0.14251848, 0.91090492, 0.91090492]],
[[ 0.14217992, 0.14217992, 0.14217992],
[ 0.14217992, 0.41549218, 0.41549218]]]]) # note the repetition, because in maxloc you're repeatedly asking for layer 0 sometimes, and sometimes repeatedly for layer 1

Numpy indexing 3-dimensional array into 2-dimensional array

I have a three-dimensional array of the following structure:
x = np.array([[[1,2],
[3,4]],
[[5,6],
[7,8]]], dtype=np.double)
Additionally, I have an index array
idx = np.array([[0,1],[1,3]], dtype=np.int)
Each row of idx defines the row/column indices for the placement of each sub-array along the 0 axis in x into a two-dimensional array K that is initialized as
K = np.zeros((4,4), dtype=np.double)
I would like to use fancy indexing/broadcasting to performing the indexing without a for loop. I currently do it this way:
for i, id in enumerate(idx):
idx_grid = np.ix_(id,id)
K[idx_grid] += x[i]
Such that the result is:
>>> K = array([[ 1., 2., 0., 0.],
[ 3., 9., 0., 6.],
[ 0., 0., 0., 0.],
[ 0., 7., 0., 8.]])
Is this possible to do with fancy indexing?

Here's one alternative way. With x, idx and K defined as in your question:
indices = (idx[:,None] + K.shape[1]*idx).ravel('f')
np.add.at(K.ravel(), indices, x.ravel())
Then we have:
>>> K
array([[ 1., 2., 0., 0.],
[ 3., 9., 0., 6.],
[ 0., 0., 0., 0.],
[ 0., 7., 0., 8.]])
To perform unbuffered inplace addition on NumPy arrays you need to use np.add.at (to avoid using += in a for loop).
However, it's slightly probelmatic to pass a list of 2D index arrays, and corresponding arrays to add at these indices, to np.add.at. This is because the function interprets these lists of arrays as higher-dimensional arrays and IndexErrors are raised.
It's much simpler to pass in 1D arrays. You can temporarily ravel K and x to give you a 1D array of zeros and a 1D array of values to add to those zeros. The only fiddly part is constructing a corresponding 1D array of indices from idx at which to add the values. This can be done via broadcasting with arithmetical operators and then ravelling, as shown above.

The intended operation is one of an accumulation of values from x into places indexed by idx. You could think of those idx places as bins of a histogram data and the x values as the weights that you need to accumulate for those bins. Now, to perform such a binning operation, np.bincount could be used. Here's one such implementation with it -
# Get size info of expected output
N = idx.max()+1
# Extend idx to cover two axes, equivalent to `np.ix_`
idx1 = idx[:,None,:] + N*idx[:,:,None]
# "Accumulate" values from x into places indexed by idx1
K = np.bincount(idx1.ravel(),x.ravel()).reshape(N,N)
Runtime tests -
1) Create inputs:
In [361]: # Create x and idx, with idx having unique elements in each row of idx,
...: # as otherwise the intended operation is not clear
...:
...: nrows = 100
...: max_idx = 100
...: ncols_idx = 2
...:
...: x = np.random.rand(nrows,ncols_idx,ncols_idx)
...: idx = np.random.randint(0,max_idx,(nrows,ncols_idx))
...:
...: valid_mask = ~np.any(np.diff(np.sort(idx,axis=1),axis=1)==0,axis=1)
...:
...: x = x[valid_mask]
...: idx = idx[valid_mask]
...:
2) Define functions:
In [362]: # Define the original and proposed (bincount based) approaches
...:
...: def org_approach(x,idx):
...: N = idx.max()+1
...: K = np.zeros((N,N), dtype=np.double)
...: for i, id in enumerate(idx):
...: idx_grid = np.ix_(id,id)
...: K[idx_grid] += x[i]
...: return K
...:
...:
...: def bincount_approach(x,idx):
...: N = idx.max()+1
...: idx1 = idx[:,None,:] + N*idx[:,:,None]
...: return np.bincount(idx1.ravel(),x.ravel()).reshape(N,N)
...:
3) Finally time them:
In [363]: %timeit org_approach(x,idx)
100 loops, best of 3: 2.13 ms per loop
In [364]: %timeit bincount_approach(x,idx)
10000 loops, best of 3: 32 µs per loop

I do not think it is efficiently possible, since you have += in the loop. This means, you would have to "blow up" your array idx by one dimension and reduce it again by utilizing np.sum(x[...], axis=...).
A minor optimization would be:
import numpy as np
xx = np.array([[[1, 2],
[3, 4]],
[[5, 6],
[7, 8]]], dtype=np.double)
idx = np.array([[0, 1], [1, 3]], dtype=np.int)
K0, K1 = np.zeros((4, 4), dtype=np.double), np.zeros((4, 4), dtype=np.double)
for k, i in enumerate(idx):
idx_grid = np.ix_(i, i)
K0[idx_grid] += xx[k]
for x, i in zip(xx, idx):
K1[np.ix_(i, i)] += x
print("K1 == K0:", np.allclose(K1, K0)) # prints: K1 == K0: True
PS: Do not use id as a variable name, since it is a Python keyword.