I have not checked if this training algorithm really optimize the object function yet in a experiment. So, mathematical formulation might be wrong.
I have not checked if this training algorithm really optimize the object function yet in a experiment. So, mathematical formulation might be wrong.