0
$\begingroup$

I need to predict some continuous data. How can I add knowledge about a range to the model? For example: dataset about morphology to predict value of red bloodcells. From knowledge from doctors or books we know that for example value cannot be negative and that value are in some allowable range, how to give that knowledge to a model?

$\endgroup$
1
  • $\begingroup$ change the output layer activation function to match the distribution you are targetting $\endgroup$
    – anon
    Aug 29, 2022 at 11:33

1 Answer 1

1
$\begingroup$

You're looking for Bayesian statistics. I'd suggest learning a PPL like Stan or Turing.jl, and using that to model your problem.

$\endgroup$

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.