Fenchel duality of Cox partial likelihood with an application in survival kernel learning

作者：

Highlights：

• The convex conjugate function of the Cox loss function based on Fenchel duality is provided for the first time.

• The derived dual form suggests an efficient algorithm for solving the kernel learning problem with censored survival outcomes.

• The optimization can be applied to many AI projects such as in the integration of multi-source molecular data for prognostic biomarker discovery in cancer studies.

摘要

•The convex conjugate function of the Cox loss function based on Fenchel duality is provided for the first time.•The derived dual form suggests an efficient algorithm for solving the kernel learning problem with censored survival outcomes.•The optimization can be applied to many AI projects such as in the integration of multi-source molecular data for prognostic biomarker discovery in cancer studies.

论文关键词：Convex conjugate,Cox model,Convex optimization,Multiple kernel learning,Fenchel dual,Survival data

论文评审过程：Received 19 May 2020, Revised 14 April 2021, Accepted 19 April 2021, Available online 24 April 2021, Version of Record 3 May 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102077