I was reading in a On the Decision Boundary of Deep Neural Networks that the final layer of a MLP can be equated to an SVM and can generate decision boundaries similar to methods with SVM. I was wondering if using this boundary detection method or another can you quantify how much probability a model assigns to each bin where a bin is a class. So for example, if a project of an input before the final layer has a SVM margin of let's say 4 from the boundaries and classification 1 and 2, can we determine how much probability it'll give class 1 or 2 after the final layer?


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.