Using Scikit-survival tree criterion differently for uncensored and censored data

[rizkiduwinanto]

I am currently making a my master thesis and would like to extend the survival tree into AFT and eventually random forest. But, I would like to ask if it is possible to implement a different criterion for y_uncensored and y_censored data in this class? I want to add them so y_uncensored follow a pdf function but y_censored a integral of cdf function from known bound to upper bound and i will add to minimize the aft loss and split https://github.com/sebp/scikit-survival/blob/master/sksurv/tree/_criterion.pyx or do i need to change the scikit library as well?

[sebp]

That's correct, if you would like to change the criteria that's used to grow trees, you need to implement scikit-learn's Criterion in Cython.