In the optimization process, obviously the precision changes by changing the number of training points. For a special case, I need to make the precision independent from the test repeating number (or the number of training points here). Can anyone give me an idea or introduce me to any other parameters that can be replaced by precision?