Change all elements in Numpy array that end up as individual arrays to floats Python

Question

I have an array in which all of the elements in the array are also arrays:

a = [[array([ 4.15167139]) array([ 2.80260218]) array([ 3.34189103])
  array([ 3.73434426]) array([ 3.76504973]) array([ 3.91946708])
  array([ 2.850741]) array([ 1.9985034]) array([ 4.05191836])
  array([ 3.46145848]) array([ 2.99489689]) array([ 2.60462582])
  array([ 1.91031189]) array([ 2.90006744]) array([ 3.69799451])
  array([ 3.83314665]) array([ 4.42917628]) array([ 5.17647658])
  array([ 4.63462677]) array([ 4.69085313]) array([ 4.84746095])
  array([ 5.04396694]) array([ 5.10152712]) array([ 3.33442499])
  array([ 4.87380637]) ...]]

I want this to be the same shape array that it is now (20,96) but instead of all of the elements within this array being arrays I would just like them to be floats:

 a = [[ 4.15167139  2.80260218  3.34189103...4.87380637 ...]]

What is the best way to iterate over this array and change all of the elements from individual arrays to floats?

Try a mix of indexing to remove extra [], and concatenate (or stack) to join arrays. — hpaulj
– hpaulj, Commented Jul 5, 2017 at 5:40
how was this array created? I think modifying its creation directly might be easier... and probably cleaner. — Julien
– Julien, Commented Jul 5, 2017 at 5:42
However you made this array, you did something wrong. You should address the problem at the source. — user2357112
– user2357112, Commented Jul 5, 2017 at 5:52
Sometimes lists of arrays, or object arrays of arrays are produced by other modules, and you have to make the best of it. — hpaulj
– hpaulj, Commented Jul 5, 2017 at 5:54

hpaulj · Accepted Answer · 2017-07-05 05:50:22Z

1

Use indexing to remove a top layer of []

In [159]: array=np.array
In [161]: a = [[array([ 4.15167139]), array([ 2.80260218]), array([ 3.34189103]),
     ...:   array([ 3.73434426]), array([ 3.76504973]), array([ 3.91946708])]]
In [162]: len(a)
Out[162]: 1
In [163]: len(a[0])
Out[163]: 6

Use concatenate or stack to join a list of arrays.

In [164]: np.stack(a[0])
Out[164]: 
array([[ 4.15167139],
       [ 2.80260218],
       [ 3.34189103],
       [ 3.73434426],
       [ 3.76504973],
       [ 3.91946708]])
In [165]: np.concatenate(a[0])
Out[165]: 
array([ 4.15167139,  2.80260218,  3.34189103,  3.73434426,  3.76504973,
        3.91946708])

Reshape as needed.

In this case np.array also works, but I found that the concatenate approach is more reliable. Sometimes np.array just produces an object array.

In [166]: np.array(a[0])
Out[166]: 
array([[ 4.15167139],
       [ 2.80260218],
       [ 3.34189103],
       [ 3.73434426],
       [ 3.76504973],
       [ 3.91946708]])

edited Jul 5, 2017 at 5:50

answered Jul 5, 2017 at 5:46

hpaulj

233k14 gold badges260 silver badges392 bronze badges

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

2 Comments

Rahul Over a year ago

In[161] for ipython 3.5 (windows) gives error, TypeError: array() argument 1 must be a unicode character, not list. How to create such array?

hpaulj Over a year ago

I had to define an alias; array = np.array. Notice I also added commas.

JohanL · Accepted Answer · 2017-07-05 05:58:36Z

1

It may be that np.concatenate is more reliable as said in hpaulj's answer, but in the current case np.array makes for a simpler solution, as it retains the dimensions. Applying it directly at the current array, will add a third dimension, however. This can be removed by slicing as:

b = np.array(a)[:,:,0]

It does not have to be more complicated than that.

answered Jul 5, 2017 at 5:58

JohanL

6,9211 gold badge16 silver badges32 bronze badges

Collectives™ on Stack Overflow

Change all elements in Numpy array that end up as individual arrays to floats Python

2 Answers 2

2 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related