Where is the place for calculating the score function?

[ccylance]

problem: The official website demonstrates how to use the l2 score function to find the best split point. However, when I tried to verify whether my understanding of the formula was correct with a small dataset, I found that the result I got was different from the result shown by the model. So I would like to ask where is the place for calculating the score function used to find the best split point?
I will be very grateful for your reply.

catboost version: catboost 1.2
Operating System: macos 14.4.1
CPU: M1
GPU: no

[ek-ak]

Hello!
These are the places in the code that you are asking for, but I'm not sure if it can be very useful. You can ask your question here.
https://github.com/catboost/catboost/blob/master/catboost/private/libs/algo/score_calcers.h#L76
https://github.com/catboost/catboost/blob/master/catboost/private/libs/algo/leafwise_scoring.cpp#L477

[ccylance]

@ek-ak Really appreciate for your reply! The l2 score function in the score_calcers.h is calculated by Scores[splitIdx] += leafApprox * leafStats.SumWeightedDelta; which is different with the formula in the Types of score functions