Numpy 3d arrays to matlab

Hi,

I have a dataset made of 3d arrays in numpy. I want to use this dataset in Matlab. I can write these 3d arrays to csv but they are written as string. I know how to deserialize them in python but not in Matlab. Is there a way to do so? Maybe using csv as an intermediate is not the best choice. Is there anything better you would recommend ?

Here is the code I use in Python for deserialization :

x_array = inputs.apply(lambda x: 
                           np.fromstring(
                               x.replace('\n','')
                                .replace('[','')
                                .replace(']','')
                                .replace('  ',' '), sep=' ')
                                .reshape(20,20)).apply(lambda x:
                                                      np.array([x]))

Thank you!

Answer 1

[SC]

Hey, These links might help you:

https://stackoverflow.com/questions/10997254/converting-numpy-arrays-to-matlab-and-vice-versa

https://stackoverflow.com/questions/31092323/how-to-exchange-multidimensional-arrays-from-matlab-to-python-numpy-and-back

https://www.mathworks.com/matlabcentral/answers/475631-in-matlab-how-can-i-write-out-a-multidimensional-array-as-a-string-that-looks-like-a-raw-numpy-array

https://bic-berkeley.github.io/psych-214-fall-2016/index_reshape.html

[Rémi Grisot]

Thank you for the links, it helped me for some parts of the solutions. Here is what I produced :

import scipy.io

arr_train = np.stack(data['hist'].values)
scipy.io.savemat('data/data.mat', mdict={'train': arr_train})
arr_valid = np.stack(data_valid['hist'].values)
scipy.io.savemat('data/data_valid.mat', mdict={'test': arr_valid})