Reproduce “An Introduction to Deep Survival Analysis Models for Predicting Time-to-Event Outcomes” 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
George H. Chen
George H. Chen's monograph from Carnegie Mellon University offers a structured introduction to deep survival analysis, unifying a wide array of deep learning models for predicting time-to-event outcomes. It systematically categorizes methods from classical statistical approaches to modern neural network-based models, providing clarity on their interrelationships and underlying principles.
Key insights
- A consistent statistical framework for right-censored time-to-event data, leveraging the likelihood function for model training, underpins diverse deep survival models.
- Survival, hazard, and cumulative hazard functions are mathematically inter-convertible, allowing models to predict any one to characterize the conditional time-to-event distribution.
- Deep survival models can be broadly categorized into proportional hazards, non-parametric (Kaplan-Meier extensions), and neural ODE-based approaches, each with distinct flexibility and assumptions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.