Does catboost implements xgboost (extreme gradient boosting) or a simple gradient boosting?

Question

On their website they say 'gradient boosting' but it seems people here compare it to other 'xgboost' algorithm. I would like to know whether it is a real extreme gradient boosting algorithm. thanks

Answer 1

Gradient boosting is a meta-algorithm. There is no simple gradient boosting. Each boosting library uses their own unique algorithm to search for regression trees, as a result, we obtain different results.

Extreme gradient boosting is just an implementation of standard gradient boosting on decision trees from xgboost with some heuristics/regularizations to improve models quality and special scheme to learn regression trees

CatBoost is another implementation of standard gradient boosting with another set of regularizations/heuristics.

So this are different algorithms.