Sum All elements of each array in an array of arrays

Question

I have this dictionary:

import numpy as np
dict={'W1': np.array([[ 1.62434536, -0.61175641, -0.52817175], 
                     [-1.07296862,  0.86540763, -2.3015387 ]]), 
     'b1': np.array([[ 1.74481176], 
                     [-0.7612069 ]]), 
     'W2': np.array([[ 0.3190391 , -0.24937038], 
                     [ 1.46210794, -2.06014071], 
                     [-0.3224172 , -0.38405435]]), 
     'b2': np.array([[ 1.13376944], 
                     [-1.09989127], 
                     [-0.17242821]]), 
     'W3': np.array([[-0.87785842,  0.04221375,  0.58281521]]), 
     'b3': np.array([[-1.10061918]])}

I need to sum all the elements of W1, W2, W3 after squared them, each one at a time then all of the three.

I used this to extract a list with the keys W(i)

l=[v for k, v in dict.items() if 'W' in k]

How could I get the sum of the squared elements in each array? When taken each array separately I do:

 np.sum(np.square(l[0]) to get 10.4889815722 for l[0]

I don't know though how to sum them all in one shot

I'm a bit confused how you get these numbers. Which elements add up to 10.4889815722 for example? — MSeifert
– MSeifert, Commented Aug 20, 2017 at 15:16
Sorry, my bad, I forget to say that I do square the elements before adding them up for 10.4889815722 that's np.sum(np.square(l[0]). Question edited to reflect that. — Abdellah Didi
– Abdellah Didi, Commented Aug 20, 2017 at 15:22

MSeifert · Accepted Answer · 2017-08-20 15:27:55Z

1

You could simply extract the sum of all values with a dictionary comprehension:

>>> res = {key: np.square(arr).sum() for key, arr in dct.items()}  # you could also use if 'W' in key here too.
>>> res
{'W1': 10.48898156439229,
 'W2': 6.7973615015702658,
 'W3': 1.1120909752613031,
 'b1': 3.6238040224419072,
 'b2': 2.5249254365039309,
 'b3': 1.2113625793838725}

Given that dictionaries are unordered having the result as a dictionary is probably better (accessible with res['W1'] for example) because otherwise the list elements would be in arbitrary order (or you need to sort the keys before putting them in a list).

To sum all W* values:

>>> sum(v for k, v in res.items() if 'W' in k)  # normal sum this time but would also work with np.sum!
18.398434041223858

answered Aug 20, 2017 at 15:27

MSeifert

154k41 gold badges356 silver badges377 bronze badges

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

Comments

Danil Speransky · Accepted Answer · 2017-08-20 15:21:27Z

0

I'm not sure I got you, but here's what I think you're looking for:

>>> import numpy as np
>>> data = [np.array([1.62434536, -0.61175641, -0.52817175]), np.array([-1.07296862, 0.86540763, -2.3015387])]
>>> [np.sum(arr) for arr in data]
[0.48441719999999999, -2.5090996900000002]

answered Aug 20, 2017 at 15:21

Danil Speransky

30.5k6 gold badges70 silver badges78 bronze badges

3 Comments

Abdellah Didi Over a year ago

That answers my question. I needed only to add np.square to get the desired result. Thank you !

F.M.F. Over a year ago

why not just np.sum(data, axis=1) in this case? this avoids the list comprehension...

NeStack Over a year ago

@F.M.F. I thought the same. But I think there is a problem when the sub-arrays, that are within the array, have different lengths. E.g. np.sum([array([ 1, 10]), array([ 1, 20])], axis=1) works well, but np.sum([array([ 1, 10, 15]), array([ 1, 20])], axis=1) throws an error, because the one subarray has 3 and the other 2 elements. Pls, let me know if you have an idea how to avoid the error

Collectives™ on Stack Overflow

Sum All elements of each array in an array of arrays

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