boosted random forest

[salmontuna1]

Is it possible to add boosted random forest feature? Similar to the xgboost feature num_parallel_tree. This will speed up the training time for the problems with a small number of features (in which using higher values of num_thread not making things faster because of overhead). If we can train couple of trees at the same time and use their averages for boosting that will make it possible to achieve similar accuracy in shorter time for the problems with small number of features. One more suggestion for the implementation, I think num_thread should only used for generating one tree just as now and the parallel trees should create its own parallel threads with a different parameter (or num_parallel_tree can used for this purpose as well) , otherwise the overhead problem emerges again, which is the case in current xgboost implementation.

Thanks,

[shiyu1994]

Thanks for the proposal. I think we can add this to the feature requests and voting hub for now. And reopen this when it is actively being worked on.

[github-actions]

This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one. Thank you for taking the time to improve LightGBM!