[FIXED] Problem execute calculations in a nested loop. TypeError: 'numpy.float64' object is not iterable

Issue

I'm trying to calculate the sum of squared errors and i'm using a nested loop.

I'm new to Python and i apologize, but i encounter the error:

  File "...", line 13, in <module>
    for y in values_subtraction_mean:
TypeError: 'numpy.float64' object is not iterable

The problem is with the second loop, when i have to calculate result in: for y in values_subtraction_mean: result = sum(math.sqrt(y))

In the second loop, it should show all values of values_subtraction_mean, so it should show 2.2, -0.8, 0.2, 1.2, -2.8. Next for each value above, a sqrt should be calculated and get 4.84, 0.64, 0.04, 1.44, 7.84. In the end you have to sum all these numbers and get 14.8

What am I doing wrong?

from numpy import mean
import math

values = [5, 2, 3, 4, 0]

mean = mean(values)

for x in values:
  values_subtraction_mean = x - mean
  print(values_subtraction_mean)
  #2.2, -0.8, 0.2, 1.2, -2.8

  for y in values_subtraction_mean:
    result = sum(math.sqrt(y))
    print(result)
    #sqrt: 4.84, 0.64, 0.04, 1.44, 7.84
    #sum and result: 14.8

I tried using this, but it doesn't solve the problem:

import numpy as np
values = np.array([5, 2, 3, 4, 0])

I tried not using numpy, calculating the mean with: sum(values) / len(values), but it doesn't work either and i get error:

TypeError: 'float' object is not iterable

Solution

You seem to have problems understanding basic python iteration, let alone numpy. So lets look at your code in detail

In [1]: import numpy as np
   ...: import math
   ...: 
   ...: values = [5, 2, 3, 4, 0]
   ...: 
   ...: mean = np.mean(values)
   ...: 
   ...: for x in values:
   ...:   values_subtraction_mean = x - mean
   ...:   print(values_subtraction_mean)
   ...: 
   ...:   for y in values_subtraction_mean:
   ...:     result = sum(math.sqrt(y))
   ...:     print(result)
   ...: 
2.2
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[1], line 12
      9 values_subtraction_mean = x - mean
     10 print(values_subtraction_mean)
---> 12 for y in values_subtraction_mean:
     13   result = sum(math.sqrt(y))
     14   print(result)

TypeError: 'numpy.float64' object is not iterable

So the error is in that for y in values...: line. The for x in values: works because values is a list. Inside that loop it calculates the mean (with a numpy mean function):

In [2]: mean
Out[2]: 2.8

In [3]: type(mean)
Out[3]: numpy.float64

In [4]: values_subtraction_mean
Out[4]: 2.2

In [5]: x
Out[5]: 5

You substracted the mean from one element of the list, and then tried to iterate on that value. The y loop is nested inside the x loop. Plus the x loop isn't accumulating any of those x-mean values.

The following line will have problems as well

sum(math.sqrt(y))

math functions only work on scalars, not lists or arrays. But python sum requires an iterable (e.g. a list). So I don't quite get what you intend here.

Corrected iteration - sort of

The standard way to iterate on a list is to append the results to a list. Changing your code:

In [9]: values = [5, 2, 3, 4, 0]
   ...: 
   ...: mean = np.mean(values)
   ...: values_sublist = []
   ...: for x in values:
   ...:   values_sublist.append(x - mean)
   ...: print(values_sublist)
   ...: result = []
   ...: for y in values_sublist:
   ...:     result.append(math.sqrt(y))
   ...: print(result)
[2.2, -0.7999999999999998, 0.20000000000000018, 1.2000000000000002, -2.8]
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[9], line 10
      8 result = []
      9 for y in values_sublist:
---> 10     result.append(math.sqrt(y))
     11 print(result)

ValueError: math domain error

values_sublist is all the list values minus their mean. So the y loop works. But it runs into another problem - it can't take the sqrt of a negative number!

numpy version

In [10]: values
Out[10]: [5, 2, 3, 4, 0]

In [11]: mean=np.mean(values)

In [12]: mean
Out[12]: 2.8

In [13]: np.array(values)-mean
Out[13]: array([ 2.2, -0.8,  0.2,  1.2, -2.8])

In [14]: np.sqrt(np.array(values)-mean)
C:\Users\14256\AppData\Local\Temp\ipykernel_8020\509743582.py:1: RuntimeWarning: invalid value encountered in sqrt
  np.sqrt(np.array(values)-mean)
Out[14]: array([1.4832397 ,        nan, 0.4472136 , 1.09544512,        nan])

It's possible to do this without the python loops, if values is a numpy array. Now the negative values produce a nan value, and a warning. There's no point to doing the sum on the array with nan.

complex values

We can deal with the sqrt issue by making values a complex dtype array:

In [15]: values_c = np.array(values,complex); values_c
Out[15]: array([5.+0.j, 2.+0.j, 3.+0.j, 4.+0.j, 0.+0.j])

In [16]: np.mean(values_c)
Out[16]: (2.8000000000000003+0j)

In [17]: values_c - np.mean(values_c)
Out[17]: array([ 2.2+0.j, -0.8+0.j,  0.2+0.j,  1.2+0.j, -2.8+0.j])

In [18]: np.sqrt(values_c - np.mean(values_c))
Out[18]: 
array([1.4832397 +0.j        , 0.        +0.89442719j,
       0.4472136 +0.j        , 1.09544512+0.j        ,
       0.        +1.67332005j])

In [19]: np.sum(_)
Out[19]: (3.0258984079294224+2.5677472440680673j)

Though I'm not sure that's what you need.

square instead of sqrt

But wait, the "sqrt" values that you want are actually the square values, not square root.

In [20]: values = [5, 2, 3, 4, 0]
    ...: mean = np.mean(values)
    ...: values_sublist = []
    ...: for x in values:
    ...:   values_sublist.append(x - mean)
    ...: print(values_sublist)
    ...: result = []
    ...: for y in values_sublist:
    ...:     result.append(y**2)
    ...: print(result); print(sum(result))
[2.2, -0.7999999999999998, 0.20000000000000018, 1.2000000000000002, -2.8]
[4.840000000000001, 0.6399999999999997, 0.04000000000000007, 1.4400000000000004, 7.839999999999999]
14.8

Now your two loops work just fine.

And written as one numpy expression:

In [21]: np.sum((np.array(values)-np.mean(values))**2)
Out[21]: 14.8

We could also use a list comprehension in place of the for loops - though we still have to use np.mean, or a loop equivalent. So it isn't a pure list calculation:

In [22]: values
Out[22]: [5, 2, 3, 4, 0]

In [23]: mean
Out[23]: 2.8

In [24]: sum((x-mean)**2 for x in values)
Out[24]: 14.8

That sqrt and complex code may be a distraction, but I'll leave that in because I think it may be instructive. It's part of debugging your code.

Answered By - hpaulj