Reproduce “Survival regression with accelerated failure time model in XGBoost” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Avinash Barnwal, Hyunsu Cho, Toby Dylan Hocking
Researchers implemented the Accelerated Failure Time (AFT) model within the XGBoost framework, enabling direct survival time prediction and comprehensive support for various censoring types. This integration delivered comparable or superior generalization performance to baselines, alongside a 6.1x to 6.7x speedup on a 20 million sample dataset when leveraging NVIDIA GPUs.
Key insights
- Accelerated Failure Time (AFT) models can be effectively integrated into scalable gradient boosting frameworks like XGBoost to support comprehensive censoring types.
- Combining the AFT model's direct time-to-event prediction with XGBoost's nonlinear modeling capabilities leads to competitive or superior generalization performance on diverse datasets.
- The AFT implementation within XGBoost significantly benefits from hardware acceleration, making it suitable for large-scale survival analysis tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.