Hi,
In param.h, CalcGainGivenWeight only concern about L2 regularization. CalcGain take L1 into consideration. But in split_evaluator.h, why CalcSplitGain use CalcGainGivenWeight rather than CalGain to calculate split gain? I thought the use of CalcGainGivenWeight not take the L1 regularization into consideration.
thank you!
