Hi, I am slightly confused and would like to know the difference between a softmax output and a multi label output given they would have same number of units depends on the number of labels you want to classify
Hi, @Nnaemeka_Nwankwo !
Multi-label output refers to a generalization of multiclass classification which involves predicting mutually non-exclusive class labels.
With softmax, when one logit (the unnormalized output of a model) is higher pushes the other logits to be smaller. Intuitively, it does the opposite of what you would want to achive in a multi-label problem, which is actually having several output logits high (meaning high probability)
Thank you Alvaro, I now have better understanding since they are “mutually non-exclusive” so each will have a high probability if it’s true. Thanks.