0
$\begingroup$

I have data that one row has input of size 1x2 (two values) and output is matrix of size 15x3 (fifteen rows with three values) like that

            |-----------------|
            |y01,2 y01,2 y01,3|
            |y02,1 y02,2 y02,3|
            |.                |
            |.                |
            |.                |
            |.                |
            |.                |
-------     |.                |
|x1,x2|     |.                |
-------     |.                |
            |.                |
            |.                |
            |.                |
            |.                |
            |y15,1 y15,2 y15,3|
            |-----------------|

This is one row and i have around 3000 inputs like that. Most values are real numbers except one output column that is binary (So output y$_{n,3}$ values are 0 or 1).

1) How should i preprocess this kind of data?

2) Is it of any sense to train two models, one for for real numbers only and the second one for binary output?

3) What neural network architecture would work best for this task?

Thanks for help.

$\endgroup$
2
  • $\begingroup$ So you are looking to understand how your two covariates in $x$ can predict the 45 response variables in $y$? $\endgroup$
    – ERT
    Jul 11, 2018 at 22:02
  • $\begingroup$ Yes, i am wondering which neural network would process it best and how should i preprocess this data, if standard zero variance and unit mean is enough $\endgroup$ Jul 12, 2018 at 10:15

1 Answer 1

1
$\begingroup$

It's hard to say what solution will work better without knowing nature of the data, but I can suggest you to try these two options:

  1. If values in the output are not related you can just reshape your output to a vector that has 45 values and build a regular neural network that takes vectors with two values as an input and outputs 45 variables

  2. If values are somehow related, like pixels in the image, than you can apply a few transformations to the input and scale it to maybe 9 variables, reshape it to the shape like 9x1 or 3x3 and apply a few convolutional operations with some extra padding in order to upscale your output to the shape 15x3. It's essentially the same thing as decoder in the convolutional autoencoder. You can check this type of autoencoder here: https://github.com/itdxer/neupy/blob/master/examples/autoencoder/stacked_conv_autoencoders.py#L52-L62

$\endgroup$
2
  • $\begingroup$ It is hard to say if data on output is related, these are properties of psychical elementary particles. $\endgroup$ Jul 16, 2018 at 16:05
  • 1
    $\begingroup$ Then you just need to try both to see what works better for your problem $\endgroup$
    – itdxer
    Jul 16, 2018 at 19:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.