XGB is a distributed and greedy gradient boosting library in kdb+/q designed to be highly efficient . It is a native implementation in kdb+/q and does not depende on other module. It implements machine learning algorithms under the Gradient Boosting framework. XGB provides a native distributed tree boosting that solve many data science problems in a fast and accurate way. You can use xgb as long as the data is in memory or parted on disk.
Why do you reimplement an existing framework in kdb+/q and not just use xgboost from Python/R inside kdb+/q?
Python/R require that the data fits into memory. The data size is the limitation. As long as the data is parted on disk you can apply xgb to your dataset. I am using it to participate the kaggle bosch competition.
- Tianqi Chen and Carlos Guestrin. XGBoost: A Scalable Tree Boosting System. In 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, 2016