Add some values at some range in numpy array?

Question

I have an numpy array that looks like this

[1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]

Now I have an 2d array that tells the start index and the end index and the value to sum. like this

[[5, 12, 29] [8, 22, 719]]

I want to add these numbers to my initial array faster.

Currently my solution is

f= np.zeros(n+1, dtype = int)

# print(rounds)
for elem in rounds:
    print('rounds : ',elem)
    one = elem[0]
    two = elem[1]
    value_add = elem[2]
# for i in range(0,n+1):
    for i in range(one-1,two):
        f[i-1] += value_add
        # print(f)

Its very slow when matrix gets too long, Any other way to optimize ?

2nd Iteration try for optimization

import numpy as np

f= np.zeros(n+1, dtype = int)

# print(rounds)
new = np.zeros(n+1)
for elem in rounds:
    # prin  t('rounds : ',elem)
    one = elem[0]
    two = elem[1]
    value_add = elem[2]
# for i in range(0,n+1):

    new[one-1:two] += value_add

I'm not entirely clear about the 'summing' part. Am I right in assuming that the elements of the original array must be increased by a certain amount, starting at the first index and ending just before the second one? That is to say, an entry like [2, 5, 3] would result in the original changing to [1, 2, 5, 5, 5, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2]? — navneethc
– navneethc, Commented Feb 29, 2020 at 15:12

Alain T. · Accepted Answer · 2020-02-29 17:34:04Z

You could create masks and use indirection to let numpy do all the work (i.e without any loops):

import numpy as np
a   = np.array([1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2])
iv  = np.array([[5, 12, 29], [8, 22, 719]])

valueMask     = (np.arange(a.size)>=iv[:,0:1]) & (np.arange(a.size)<iv[:,1:2])
valueIndex    = np.max((np.arange(iv.shape[0])+1)[:,None]*valueMask,axis=0)-1
targetMask    = valueIndex>=0
a[targetMask] = iv[valueIndex[targetMask],2]

output:

print(a)

[  1   2   2   2   2  29  29  29 719 719 719 719 719 719 719 719 719 719  719]

The valueMask variable is a true/false matrix of positions in a where each value of iv would be assigned.

The valueIndex variable converts the masks into a 1D array indicating which of the iv value (index) will be assigned last at each position. It will contain -1 for positions that are not changed.

The targetMask variable picks out the positions of a that will actually receive a new value and excludes the positions that will not be changed. It results in a true/false mask of positions to use in the indirection for value assignment.

The last step is to assign the subset of elements in a defined by targetMask with the corresponding values of iv using the last value (highest index of iv) that would be assigned.

I hope this will give you the performance you need. I measured assigning 300 ranges to a 2000 item array and got a time of 0.0028 seconds on my laptop.

navneethc · Accepted Answer · 2020-02-29 15:20:38Z

0

How about this?

a = np.array([1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2])
instr_array = np.array([[5, 12, 29], [8, 22, 719]])

for start, end, incr in instr_array:
    a[start:end] += incr

answered Feb 29, 2020 at 15:20

navneethc

1,4761 gold badge11 silver badges22 bronze badges

3 Comments

Abhay Over a year ago

yeah and I still getting time penalty for 8 out of 14 test cases.

navneethc Over a year ago

Where are these tests hosted, if I may ask?

Abhay Over a year ago

On hackerrank plateform.

Collectives™ on Stack Overflow

Add some values at some range in numpy array?

2 Answers 2

Comments

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related