Two new research papers propose enhancements to the XGBoost algorithm for improved regression and prediction. The first paper introduces QXGB, a quantile-based gradient boosting framework that uses a convolution smoothed loss to better estimate conditional quantiles and characterize extreme outcomes, demonstrated in a PM2.5 prediction task. The second paper addresses the robustness of XGBoost against outliers by exploring alternative loss functions, proposing MM-XGBoost as a two-step procedure that balances robustness with prediction accuracy. AI
IMPACT These research papers offer advancements in predictive modeling and robustness for machine learning algorithms, potentially improving accuracy and reliability in various applications.
RANK_REASON Two academic papers published on arXiv proposing novel methods for the XGBoost algorithm.
- arXiv
- Huber loss function
- M-estimators
- MM-XGBoost
- S-estimators
- tau-estimators
- XGBoost
- California
- PM$_{2.5}$
- QXGB
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
