0
$\begingroup$

I want to make a multilabel classifier where the labels are dependent on one another. Concretely, I have a situation where:

If label A is true, then label B is also true.

If label B is true, label A could be true or false.

If label B is false, then label A is false.

How can I model this in a multilabel classifier?

$\endgroup$
1
$\begingroup$

In this case the easiest way is to use a simple single-label classifier with three labels:

  • A0B1
  • A1B1
  • A0B0

As long as the number of combinations is reasonable, that's what will give the best results while satisfying the constraints.

By the way "If label B is false, then label A is false" is logically implied by "if label A is true, then label B is also true". (contrapositive).

$\endgroup$

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.