I want to do supervised learning. I have a labeled dataset. Each label is normally a real number between [0,1] corresponding to a location of a certain object in the data, or NAN if the object is not present in the data.
Is there a "recommended way" to convert my labels into valid labels?
My ideas are:
- Train two different classifiers, one for existence of object, and one for its location
- Train one classifier, but with two outputs, one binary for existence, and one float for location
- Relabel all NANs as -1, and train one classifier that can be -1 or in [0,1]