I have a classification problem where I want to want to use probabilities instead of classes to train my model to learn to output probabilities. In my dataset, I have instances where the probabilities of two classes are almost equal and I would like the model to be able to learn these subtleties instead of me just providing the class for each instance. Is there an ML model that can handle this?


  • 1
    $\begingroup$ The standard cross entropy formulation works fine mathematically. If the implementation you use for the loss allows for soft labels, then this should work computationally as well. $\endgroup$ – GeoMatt22 Feb 24 at 5:56

Your Answer

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

Browse other questions tagged or ask your own question.