> ML_LITERATURE // CHEN-GUESTRIN-2016-XGBOOST-SCALABLE-TREE-BOOSTING_v1.0
XGBoost: A Scalable Tree Boosting System
Tianqi Chen, Carlos Guestrin · ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) (2016)
systems2016industry-standardthirdPartyReproduced
Principal Contribution
Second-order Taylor expansion tree loss optimization with column subsampling, cache-aware block structure, and out-of-core computing.
Operational Relevance
The workhorse system for production tabular classification, fraud detection, risk scoring, and Kaggle competition victories.
Assumptions
- Second-order Hessian gradients accurately capture curvature for optimal leaf weight scoring
Limitations
- Exact greedy split finding does not scale to high cardinality categorical features without one-hot encoding or binning
Connected Algorithms, Architectures & Tools
Related Algorithms:
Related Architectures:
Implementing Libraries: