I am new to xgboost and not so good at math.

I want to use a self defined objective function, here is the expression:

obj = (y * y_pred).sum() / (y_pred).abs().sum()

y is a n-dimensional vector.

I have some experience for tensorflow, in tensorflow, I just need to define the function, it will calculate the derivatives for me automatically.

But in xgboost, the customized obj function needs to return the first and second derivatives.

I dont know how to take derivatives for abs(). Is there good method can take derivatives? Can you help on this?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.