The functions ClassifierMeasurements and NetMeasurements expect objects that are created through Classify or are objects with head NetGraph, NetChain etc.
I've got a classifier that I've created through my own algorithm and I would like to use ClassifierMeasurements to easily get assessment tools like confusion matrix, ROC, and all the other statistics that ClassifierMeasurements can compute. However, if I provide my function to ClassifierMeasurements, it complains that my function is not an adequate classifier object.
The only solution that I can see right now, is to find a third party implementation or implement the statistics that I need by myself.
My question is:
Is there a way to wrap my classifier such that ClassifierMeasurements would work with it?
PS: Of course, I could adapt my classifier that it either gives a probability vector, or the class index or anything else as output.
