xgboost/README.md
2014-03-26 16:25:44 -07:00

41 lines
1.2 KiB
Markdown

xgboost: eXtreme Gradient Boosting
=======
A General purpose gradient boosting (tree) library.
Authors:
* Tianqi Chen, project creater
* Kailong Chen, contributes regression module
Turorial and Documentation: https://github.com/tqchen/xgboost/wiki
Features
=======
* Sparse feature format:
- Sparse feature format allows easy handling of missing values, and improve computation efficiency.
* Push the limit on single machine:
- Efficient implementation that optimizes memory and computation.
* Layout of gradient boosting algorithm to support generic tasks, see project wiki.
Supported key components
=======
* Gradient boosting models:
- regression tree (GBRT)
- linear model/lasso
* Objectives to support tasks:
- regression
- classification
* OpenMP implementation
Planned components
=======
* More objective to support tasks:
- ranking
- matrix factorization
- structured prediction
File extension convention
=======
* .h are interface, utils and data structures, with detailed comment;
* .cpp are implementations that will be compiled, with less comment;
* .hpp are implementations that will be included by .cpp, with less comment