arXiv 2511.095672025-11-23cs.LG

让专家发言：通过专家混合头改进生存预测与校准

Let the Experts Speak: Improving Survival Prediction & Calibration via Mixture-of-Experts Heads

Department of Computer Science Columbia University USA（哥伦比亚大学计算机科学系）；Department of Computer Science New York University USA（纽约大学计算机科学系）；Department of Neurology Columbia University Medical Center USA（哥伦比亚大学医学中心神经病学系）；Department of Computer Science Barnard College USA（巴纳德学院计算机科学系）；Department of Biomedical Informatics Columbia University Medical Center USA（哥伦比亚大学医学中心生物医学信息学系）；NewYork-Presbyterian Hospital at Columbia University Medical Center USA（哥伦比亚大学医学中心新英格兰-纽约 Presbyterian 医院）

Todd Morrill, Aahlad Puli, Murad Megjhani, Soojin Park, Richard Zemel

备注：: Accepted as a proceedings paper at the 2025 Machine Learning for Health Symposium and as a workshop paper at the Learning from Time Series for Health workshop at NeurIPS 2025

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出了一种改进生存预测和校准的混合专家架构，通过更具表现力的专家实现更准确的患者分组和预测。

摘要：

深度混合专家模型在生存分析问题中受到了广泛关注，特别是由于它们能够将相似患者聚类在一起。在实践中，分组往往以牺牲关键指标如校准误差和预测准确性为代价。这是由于混合专家模型所施加的限制性归纳偏差，即对个体患者的预测必须看起来像所分配群体的预测。我们能否发现存在的患者群体结构，同时提高校准和预测准确性？在本工作中，我们介绍了几种离散时间深度混合专家（MoE）基于架构用于生存分析问题，其中一种实现了所有需求：聚类、校准和预测准确性。我们发现，这一系列MoE之间的一个关键区别在于专家的表达能力如何。我们发现，能够根据患者定制预测的更具表现力的专家优于依赖固定群体原型的专家。

英文摘要：

Deep mixture-of-experts models have attracted a lot of attention for survival analysis problems, particularly for their ability to cluster similar patients together. In practice, grouping often comes at the expense of key metrics such as calibration error and predictive accuracy. This is due to the restrictive inductive bias that mixture-of-experts imposes, that predictions for individual patients must look like predictions for the group they're assigned to. Might we be able to discover patient group structure, where it exists, while improving calibration and predictive accuracy? In this work, we introduce several discrete-time deep mixture-of-experts (MoE)-based architectures for survival analysis problems, one of which achieves all desiderata: clustering, calibration, and predictive accuracy. We show that a key differentiator between this array of MoEs is how expressive their experts are. We find that more expressive experts that tailor predictions per patient outperform experts that rely on fixed group prototypes.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。