xgboost-adv-workshop-LA
xgboost-adv-workshop-LA copied to clipboard
szilard
to discuss: main points of xgboost paper on arxiv
Eq. 6 or Algo 1: what's the default gamma and lambda and how to change it?
( https://github.com/dmlc/xgboost/blob/master/doc/parameter.md )
How different is this from standard gradient boosting?
3.4 Sparsity-aware Split Finding versus NEWS file: Changes in R library: switched from 0 to NA for missing values.
(4.1) Data in each block is stored in the compressed column (CSC) format, with each column sorted by the corresponding feature value.
not in the paper, but anyways: Allreduce runtime called Rabit; Distributed XGBoost YARN on AWS; JVM/Spark
on disk: https://xgboost.readthedocs.io/en/latest/how_to/external_memory.html
