I have some data with soft labels and I am trying to figure out the best approach to solve the problem with Machine Learning (since regular classification is of the table, i.e. hard labels). However, whenever I look up "soft label" materials, I keep getting pointed to label smoothing. Is this the main/only technique to deal with soft labels?

  • $\begingroup$ Hello. Welcome to Artificial Intelligence Stack Exchange. It may be a good idea to provide more details about your problem. In fact, if by "soft labels" you mean a number between 0 and 1, in theory, I think you could still apply typically ML techniques to solve your problem (i.e. some of these produce a probability), although I have no idea of whether they will work or not. So, I am not sure why you say that "regular classification" is not applicable to your problem. You may want to elaborate on this point. In any case, the main question is ok even without the specification of your problem. $\endgroup$
    – nbro
    Aug 24, 2021 at 23:00


You must log in to answer this question.

Browse other questions tagged .