Scholarlink

This study investigates whether integrating a Variational Information Bottleneck (VIB) into deep survival models improves generalization across heterogeneous populations. We evaluate performance on preterm birth prediction using three distinct datasets, training on one and testing on the others. Compared to classical and non-VIB deep models, the VIB-enhanced models achieved higher time-dependent AUC in several cases, suggesting potential benefits despite limited sample sizes.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Generalizability of AI Survival Models in the Context of Preterm Birth Prediction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Anne Fischer,
  • Isabelle Dehaene,
  • Mark Hoogendoorn

摘要

This study investigates whether integrating a Variational Information Bottleneck (VIB) into deep survival models improves generalization across heterogeneous populations. We evaluate performance on preterm birth prediction using three distinct datasets, training on one and testing on the others. Compared to classical and non-VIB deep models, the VIB-enhanced models achieved higher time-dependent AUC in several cases, suggesting potential benefits despite limited sample sizes.