Slicing and getting the max value with Numpy Python

Question

Im trying to write a code with numpy where it outputs the maximum value between indexes. I think using argmax could be usable. However I do not know how I can use slices without using a for loop in python. If there is a pandas function for this it could be useable too. I want to make the computation as fast as possible.

list_ = np.array([9887.89, 9902.99, 9902.99, 9910.23, 9920.79, 9911.34, 9920.01, 9927.51, 9932.3, 9932.33, 9928.87, 9929.22, 9929.22, 9935.24, 9935.24, 9935.26, 9935.26, 9935.68, 9935.68, 9940.5])
indexes = np.array([0, 5, 10, 19])

Expected result:

Max number between index(0 - 5):  9920.79 at index 5
Max number between index(5 - 10): 9932.33 at index 10
Max number between index(10 - 19): 9940.5 at index 19

Ehsan · Accepted Answer · 2021-01-29 00:31:48Z

1

You can use reduceat directly yo your array without the need to splice/split it:

np.maximum.reduceat(list_,indexes[:-1])

output:

array([9932.33, 9929.22, 9940.5 ])

answered Jan 29, 2021 at 0:31

Ehsan

12.5k2 gold badges24 silver badges36 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Ananda · Accepted Answer · 2021-01-28 05:24:55Z

1

Assuming that the first (zero) index and the last index is specified in the indexes array,

import numpy as np

list_ = np.array([9887.89, 9902.99, 9902.99, 9910.23, 9920.79, 9911.34, 9920.01, 9927.51, 9932.3, 9932.33, 9928.87, 9929.22, 9929.22, 9935.24, 9935.24, 9935.26, 9935.26, 9935.68, 9935.68, 9940.5])
indexes = np.array([0, 5, 10, 19])

chunks = np.split(list_, indexes[1:-1])
print([c.max() for c in chunks])

max_ind = [c.argmax() for c in chunks]
print(max_ind + indexes[:-1])

It's not necessary that each chunk will have the same size with an arbitrary specification of indices. So The vectorization benefits of numpy is going to be lost in there one way or another (Since you can't have a numpy array where each element is of a different size in memory which also has all the benefits of vectorization).

At least one for loop is going to be necessary, I think. However, you can use split, to make the splitting a numpy-optimized operation.

edited Jan 28, 2021 at 5:24

answered Jan 28, 2021 at 4:47

Ananda

3,3106 gold badges28 silver badges49 bronze badges

4 Comments

tony selcuk Over a year ago

Thank you it works is there a way I could get the indexes of the max values as well?

Ananda Over a year ago

Indices in the original array right? Not the sub-chunks?

Ananda Over a year ago

Modified the answer to also print out the indexes.

tony selcuk Over a year ago

Thank you I will try to find a way if I could get rid of the list comprehension other than that the code is very well made thanks a lot

Collectives™ on Stack Overflow

Slicing and getting the max value with Numpy Python

2 Answers 2

Comments

4 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related