-
Notifications
You must be signed in to change notification settings - Fork 214
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add gbdt #510
Add gbdt #510
Conversation
Several modifications are added according to the discussion with @qbc2016 . @rayrayraykk To use GBDT in VFL, please specify |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Although I don't look closely at the implementation details, it looks good.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
I modified the GBDT algo, here are three comments:
(1) the only two differences with XGB in class Tree are gain computation and weight computation
(2) the tree model in GBDT dose not need hessian, but the Trainer class contains the hessian computation, for convenience, I did not modify Trainer, so I still keep the hessian in class Tree
(3) as (2) says, in class Tree_GBDT, in order to not modify the Trainer class, I copied function update_child from Tree_XGB, though in fact, Tree_GBDT does not need to compute hess