I need to input data conditionally to my deep network. In order to explain cases, I'd like to give an example. Assume that I have a 50-attribute dataset. For some attributes, a specific part of hidden layers is responsible, and for others, a different part is responsible. Also, for some cases, the same parts of the hidden layers might intersect. I think I can decide which attributes must go which hidden neurons in the input layer by using some kind of if-else block. However, I could not figure out how.

My current idea

I can enter an identity element for some attributes. For example, I have att1, att2, att3, etc. I have ins1, ins2, etc. For ins1 -> att1 = 0.5, att2 = 0.2, att3 = None For ins2 -> att1 = 0.1, att2 = None, att3 = None

But, if I do this approach, the number of attributes for an instance becomes bigger unnecessarily.

End of my current idea

Are there any opinions on this? Should I rearrange my excel file or is there any way to use if-else conditions? Regards,

  • $\begingroup$ This is very confusing, so to clarify: You have (or want?) a neural network that has that capacity to accept all 50 attributes, BUT is able to only accept a subset of them at a time? So you can input attributes [2,5,7] and nothing else? Or am I misunderstanding $\endgroup$
    – Recessive
    Jul 9, 2021 at 6:16
  • $\begingroup$ Hello. Welcome to AI SE. Could you please put your specific question in the title (to give an idea of what your question is to people that scroll down the list of questions)? $\endgroup$
    – nbro
    Jul 9, 2021 at 20:54


You must log in to answer this question.