Error: Setting an array element with a sequence with keras

Question

I am building a CNN and have been getting this error when trying to perform:

from tensorflow.keras import utils
trainY=utils.to_categorical(trainY)

ValueError: setting an array element with a sequence.

My trainY are actually labels, and it looks like this:

labels
array([list(['noise']), list(['noise']), list(['noise', 'point_source']),
       list(['noise']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise']), list(['noise']),
       list(['noise', 'point_source']), list(['noise']),
       list(['noise', 'point_source']), list(['noise', 'point_source']),
       list(['noise']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise']), list(['noise']),

Any suggestions how to fix this? Many thanks!

Nicolas Gervais · Accepted Answer · 2020-02-24 13:38:33Z

1

You can do that with sklearn.preprocessing.MultiLabelBinarizer

import numpy as np

labels = np.array([list(['noise']), list(['noise']), list(['noise', 'point_source']),
       list(['noise']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise']), list(['noise']),
       list(['noise', 'point_source']), list(['noise']),
       list(['noise', 'point_source']), list(['noise', 'point_source']),
       list(['noise']), list(['noise', 'point_source']),
       list(['noise', 'point_source']), list(['noise']), list(['noise'])])

That was what you had. Now you need to do this:

from sklearn.preprocessing import MultiLabelBinarizer

as_list = [list(i) for i in labels]

mlb = MultiLabelBinarizer()
ohe = mlb.fit_transform(as_list) # you might need to add .astype(float)

This is what you'll end up with:

array([[1, 0],
       [1, 0],
       [1, 1],
       [1, 0],
       [1, 1],
       [1, 1],
       [1, 1],
       [1, 1], ...

edited Feb 24, 2020 at 13:38

answered Feb 24, 2020 at 3:20

Nicolas Gervais

36.9k23 gold badges123 silver badges160 bronze badges

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

9 Comments

saraherceg Over a year ago

Hey Nicolas, thank you for your help! Now, when I try to do what you did I get a new error when setting my trainX and trainY split:

(trainX, testX, trainY, testY) = train_test_split(signals, labels, test_size=0.2, random_state=42)  Found input variables with inconsistent numbers of samples: [100, 154]

Does this mean they are not equal length now? Any tips on how to fix that? I am trying to d make a multi-label classifier, really appreciate your help! @NicolasGervais

Nicolas Gervais Over a year ago

Maybe remove the parentheses? Or do the train test split before using to_categorical? I'm unable to help you further. What's the error

saraherceg Over a year ago

hm even when I do to_categorical after the train/test split I now have a trainY of shape (124, 2) and before was (80, )' , just as the trainX. I also removed the parenthesis. Hm strange, the error is: Input arrays should have the same number of samples as target arrays. Found 80 input samples and 124 target samples.` Thanks again for the help!

Nicolas Gervais Over a year ago

So this is supposed to be the target for one observation or two? list(['noise', 'point_source'])

Nicolas Gervais Over a year ago

See my updated answer. It doesn't seem possible to use targets with different length with to_categorical(), and I am not able to use keras at the moment. I hope this solution is ok for you

|

Collectives™ on Stack Overflow

Error: Setting an array element with a sequence with keras

1 Answer 1

9 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

9 Comments

Your Answer

Sign up or log in

Post as a guest

Related