November 4, 2026 at 1:30 PM
Abstract:
Survival analysis aims to predict the time until an event of interest, such as death, disease progression, or equipment failure. Unlike standard regression, it must account for censoring, which occurs when the event has not been observed for some individuals, for example because they leave the study or the study ends. This makes survival prediction particularly challenging, especially on small datasets commonly encountered in clinical research.
In the first part of this talk, we will introduce the fundamental concepts of survival analysis, including the survival function, cumulative incidence function, and hazard function, which provide complementary ways of describing time-to-event outcomes. We will also discuss how survival models are evaluated, focusing on commonly used metrics such as the concordance index, the Brier score, and calibration. These concepts will provide the necessary background to understand the challenges of survival prediction.
Recent advances in tabular foundation models have demonstrated remarkable performance in classification and regression, often outperforming traditional methods such as gradient-boosted trees. However, their success has not yet translated to survival analysis, where the Cox proportional hazards model, introduced over 50 years ago, remains a strong baseline.
In the second part of the talk, we introduce SurvICL, a tabular foundation model that predicts individual survival curves through in-context learning, without dataset-specific training or hyperparameter tuning. SurvICL is pretrained entirely on synthetic datasets generated from a flexible proportional-odds prior and learns to predict survival curves using a simple quantile regression loss. We also present our benchmark of 63 survival datasets, the largest to date, and show how SurvICL outperforms 20 competing methods, including classical statistical models, tree-based approaches, deep learning methods, and other foundation models. In particular, SurvICL is the first tabular foundation model to outperform Cox on small survival datasets.
E 603, Centrale building
français
English