Staying Alive: Uncensored Survival Analysis with Tabular Foundation Models
Researchers are exploring the application of tabular foundation models (TFMs) to complex time-series prediction tasks, particularly in prognostics and health management (PHM) and survival analysis. These models, adapted for time-series data through methods like in-context learning or specific pre-training, show promise in handling fragmented and censored data efficiently. Initial results suggest TFMs can outperform traditional sequence models and even specialized survival analysis techniques, especially in low-data scenarios. AI
IMPACT Extends foundation model capabilities to censored time-series data, potentially improving predictive maintenance and healthcare analytics.