I have a dataset containing 40 input columns and 12 output columns (float values).

I supposed it's a regression problem and I am wondering of how to choose the ideal architecture of the Neural Network i didn't find examples of neural Network regression with multiple output What is the right way to do it?


Just use an output layer with 12 neurons instead of 1. Qualitatively there is no difference. For regression the output activations should be linear, and you have a few choices for cost: RMSE, MAE, or Huber Loss to name a few.

| improve this answer | |
  • $\begingroup$ Thank you for your answer, the number of outputs is equal to 12, have you tried this kind of regression problems? because I've tried and the point is that after learning almost for all inputs the outcome is similar with small changes in decimal numbers. $\endgroup$ – Media Feb 14 '18 at 5:15
  • $\begingroup$ It definitely sounds like something is broken. But very hard to say what without seeing at least your code, and preferable your data too. $\endgroup$ – Imran Feb 14 '18 at 5:23
  • $\begingroup$ I guess you have seen professor Andrew Ng's convnets tutorial. It was annotation of face that he disscussed, Actually I meant in regression tasks when the number of outputs is large, things always don't go in right way. $\endgroup$ – Media Feb 14 '18 at 5:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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