adding multiple arrays together to form one array in python

Question

I have a list (called ys) of three lists (ys[0], ys[1], ys[2]). Each of the three lists contains 10 arrays each of length 1534. The first list is shown below for example:

In [57]: print ys[0]
[array([ 1655.13816505,  1547.98589715,   572.2745519 , ...,    89.10781318,
      93.89836665,    95.07931966]), array([ 1222.2549591 ,   375.13612313,  1117.62684517, ...,    92.01444874,
      96.37158146,    98.09804547]), array([ 405.61715294,  347.45411   ,  458.31631866, ...,   95.87440348,
     92.59379305,   96.88934008]), array([  958.5300296 ,   690.68863703,  1315.69537196, ...,    96.02464434,
      94.58280479,    93.77347022]), array([ 276.97463055,  457.34617477,  908.78859867, ...,   94.75916652,
     94.02373941,   94.15538106]), array([ 1822.50632297,   596.38771818,  1163.05119636, ...,    90.92428715,
      97.46551579,    90.34230747]), array([ 1609.11576638,  1343.19751488,   891.16993616, ...,    94.081789  ,
      91.84201144,    94.06961381]), array([ 1481.02653876,   843.39342494,  1208.37885821, ...,    95.86349883,
      95.93122661,    92.94565202]), array([ 1330.1940189 ,   844.70910408,  1151.75233836, ...,    98.82465514,
     100.37876234,    96.15178672]), array([ 1287.56325832,   365.89812057,  1034.15108853, ...,    90.64446465,
      94.99436954,    90.88272168])]

I would like to find a way to sum these arrays element by element so that I am left with one array of length 1534. The excerpt from the code that generates the list of three lists is this:

def ys(norm2, ell_T):
    t=[]
    e=[]
    b=[]
    T=norm2[0]
    E=norm2[1]
    B=norm2[2]
    for i in range (0, len(T)):
        yt=((ell_T * (ell_T+1) * T[i])/(2*pi))
        ye=((ell_T * (ell_T+1) * E[i])/(2*pi))
        yb=((ell_T * (ell_T+1) * B[i])/(2*pi))
        t.append(yt)
        e.append(ye)
        b.append(yb)
    return t, e, b

ys=ys(norm2, ell_T)

My attempt at the summation and a subsequent average (which doesn't work) is shown below:

def averageys(ys, theory):
    t=[]
    e=[]
    b=[]
    for i in range(1, len(ys[0])):
        yst_arrays=ys[0][i]
        yse_arrays=ys[1][i]
        ysb_arrays=ys[2][i]
        sumT=np.add(yst_arrays)
        sumE=np.add(yse_arrays)
        sumB=np.add(ysb_arrays)
        avt=sumT/len(ys[0])
        ave=sumE/len(ys[0])
        avb=sumB/len(ys[0])
        t.append(avt)
        e.append(ave)
        b.append(avb)
    return t, e, b

averageys=averageys(ys, theory)

The problem seems to be the np.add functions. They require inputs such as np.add(ys[0][0], ys[0][1], etc...) but I want to generalise this so it isn't fixed to 10 arrays.

Your final result will be three arrays of 1534 elements each? — wwii
– wwii, Commented Dec 12, 2014 at 1:22

Marcin · Accepted Answer · 2014-12-11 23:43:48Z

2

If I correctly understand, you want to sum the arrays in e.g. ys[0] together? Thus, I made a dummy example, hopfully simulating your case

import numpy as np

ys = [[np.array([1,1,1]), np.array([1,1,1])]*5,   # 10 array of length 3 instead 1534
      [np.array([2, 2,2]), np.array([2,2,2])]*5,  # 10 array of length 3 instead 1534
      [np.array([3, 3,3]), np.array([3,3,3])]*5]  # 10 array of length 3 instead 1534

Thus, you can sum them simply by:

print(list(map(sum, ys)))

This results in:

[array([10, 10, 10]), array([20, 20, 20]), array([30, 30, 30])]

And if you want to sum all of them:

print(sum(list(map(sum, ys))))
# gives: [60 60 60]

answered Dec 11, 2014 at 23:43

Marcin

241k16 gold badges315 silver badges368 bronze badges

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

Comments

Oliver W. · Accepted Answer · 2014-12-11 23:59:11Z

Have you tried converting your list (y) of lists of numpy arrays to one single numpy array and then summing over one of the axes? Working with numpy arrays has a lot of advantages over normal python constructs (lists and all kinds of list comprehensions or functional constructs such as map) when you're working with (a lot of) numerical data.

As your code isn't a minimal working example, I'll use a different example with less arrays:

>>> f0 = [np.random.random_integers(0, 10, (3,)) for _ in range(2)]    
>>> f1 = [np.random.random_integers(0, 10, (3,)) for _ in range(2)]    
>>> f2 = [np.random.random_integers(0, 10, (3,)) for _ in range(2)]    
>>> y = [f0, f1, f2]    
>>> y 
[[array([7, 8, 3]), array([ 7, 10,  0])],
 [array([2, 7, 8]), array([2, 6, 7])],
 [array([3, 5, 0]), array([7, 3, 6])]]
>>> y[0]  # it's a list of 2 numpy arrays, similar to your example
[array([7, 8, 3]), array([ 7, 10,  0])]
>>> arr = np.array(y)  # conversion of list of lists of arrays to one numpy array
>>> arr  # it is now an array with 3 axes. Its shape is (3, 2, 3)
array([[[ 7,  8,  3],
        [ 7, 10,  0]],

       [[ 2,  7,  8],
        [ 2,  6,  7]],

       [[ 3,  5,  0],
        [ 7,  3,  6]]])

To sum only the arrays present within f0, f1 and f2, you could:

>>> arr.sum(axis=1) 
array([[14, 18,  3],
       [ 4, 13, 15],
       [10,  8,  6]])

To sum over f0, f1 and f2 (all together), you could take the sum of the previous array over axis=0 or

>>> f = arr.reshape(3*2, 3); f 
array([[ 7,  8,  3],
       [ 7, 10,  0],
       [ 2,  7,  8],
       [ 2,  6,  7],
       [ 3,  5,  0],
       [ 7,  3,  6]])

>>> f.sum(axis=0)  # take the sum over all rows
array([28, 39, 24])

Once you've summed all these, you could calculate the mean by dividing by the number of rows you've summed. For the first scenario (taking the sum only over f0, f1 and f2 separately), you could even just say arr.mean(axis=1), which will give you:

>>> arr.mean(axis=1)
array([[ 7. ,  9. ,  1.5],
       [ 2. ,  6.5,  7.5],
       [ 5. ,  4. ,  3. ]])

Which you can easily verify by hand from the example above.

Hi @JKM if this or any answer has solved your question please consider accepting it by clicking the check-mark. This indicates to the wider community that you've found a solution and gives some reputation to both the answerer and yourself. There is no obligation to do this though.

Community · Accepted Answer · 2017-05-23 12:20:32Z

1

You can sum up the individual arrays step at a time:

# Sums up the 10 arrays in each of the 3 lists
reduced_ys = [[sum(value) for value in zip(*ys_list)] for ys_list in zip(*ys)]

# Sums up the 3 remaining lists
final_ys = [sum(ys_list) for ys_list in zip(*reduced_ys)]

The syntax here can also be seen in the question Merging/adding lists in Python.

edited May 23, 2017 at 12:20

CommunityBot

11 silver badge

answered Dec 11, 2014 at 23:43

Kevin Ji

10.5k4 gold badges43 silver badges65 bronze badges

Comments

boiiocks · Accepted Answer · 2014-12-12 00:02:15Z

0

Here's your solution.

def averageys(ys):
    return list(map(sum, (list(map(sum, ys)))))

answered Dec 12, 2014 at 0:02

boiiocks

33 bronze badges

Comments

Community · Accepted Answer · 2017-05-23 11:50:56Z

0

If you just want to generalize and fix your function:

def averageys(ys, theory):
    # seperate the three groups of data
    yst, yse, ysb = ys

    # number of items in the sum
    # assumes len(yst)==len(yse)==len(ysb)
    # n needs to be a float for the mean to work
    n = float(len(yst))

    # element-wise sum of each item in the sublists
    # the * operator preceding the argument will unpack each item in the sublist
    sumT=np.add(*yst)
    sumE=np.add(*yse)
    sumB=np.add(*ysb)

    avt=sumT/n
    ave=sumE/n
    avb=sumB/n
    return avt, ave, avb

BUT: see Oliver W.'s answer for the correct way to do this.

edited May 23, 2017 at 11:50

CommunityBot

11 silver badge

answered Dec 12, 2014 at 0:56

wwii

23.9k7 gold badges42 silver badges80 bronze badges

Collectives™ on Stack Overflow

adding multiple arrays together to form one array in python

5 Answers 5

Comments

1 Comment

Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

5 Answers 5

Comments

1 Comment

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related