Nerual Network with the hinge loss

by user2729400   Last Updated March 08, 2018 08:19 AM

I was wondering on the usage of Hinge loss function in neural networks. Hinge expects a ground truth of -1 or 1. Should I use tanh or linear activation in the ANN output layer then?

Related Questions

Missclassification cost in multilayer perceptron

Updated April 07, 2017 13:19 PM

Is `sigmoid` required for binary cross entropy?

Updated December 09, 2018 00:19 AM