Censored Flow Matching for Synthetic Time-to-Event Generation
Abstract
Generating time-to-event data is increasingly important in survival analysis, but it remains statistically challenging because event times are often censored and thus prevent flexible generative modeling. To fill this gap, we propose a novel method called censored flow matching (CFM), aiming at providing a flexible flow-based generative mechanism for synthesizing time-to-event data as well as learning their corresponding probabilistic densities. Unlike existing methods, CFM does not rely on survival likelihoods to handle censoring, but instead exploits the monotonicity property of differential flows to address censoring via pullback sampling. Under this framework, we implement an alternating procedure between pullback sampling and CFM steps, leading to a flow-based expectation--maximization (EM) algorithm tailored for time-to-event inference. Under mild distributional conditions, we establish the convergence of CFM to the true event-time distribution as the EM iterations proceed. We then apply CFM to generate time-to-event data and estimate their probabilistic densities, where the observed censoring time can be incorporated to provide truncated sampling or truncated density estimation for survival inference. Through extensive simulation studies, we demonstrate the benefits of CFM for complex and large-scale data generation. Finally, we apply CFM to analyze data from the Organ Procurement and Transplantation Network (OPTN) and the Medical Information Mart for Intensive Care IV (MIMIC-IV), showcasing its benefits for synthetic and predictive survival analyses in terms of data fidelity, predictive performance, and empirical privacy characteristics.
Related
- Generalized Wasserstein Flow Matching: Nested Couplings for Point Cloud Generation
- UniFlow: Unified Multi-Scale Flow Matching for Time Series Generation
- Counterfactual Generation via Flow Matching: Coupling-Sensitive End-to-End Rates
- Differentially Private Synthetic Data via Generalized Moment Matching
- Weighted Adjoint Matching: Finetuning Flow Matching Models with Time-Dependent Control Cost
BibTeX
@misc{tan2026censored,
title = {{Censored Flow Matching for Synthetic Time-to-Event Generation}},
author = {Jianbin Tan and Xingran Chen and Kevin He and Zhenke Wu and Bhramar Mukherjee and Anru R. Zhang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.24233},
eprint = {2610.24233},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.24233},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Censored Flow Matching for Synthetic Time-to-Event Generation
AU - Jianbin Tan
AU - Xingran Chen
AU - Kevin He
AU - Zhenke Wu
AU - Bhramar Mukherjee
AU - Anru R. Zhang
PY - 2026
DA - 2026/10/16
PB - arXiv
DO - 10.48550/arxiv.2610.24233
UR - https://arxiv.org/abs/2610.24233
AB - Generating time-to-event data is increasingly important in survival analysis, but it remains statistically challenging because event times are often censored and thus prevent flexible generative modeling. To fill this gap, we propose a novel method called censored flow matching (CFM), aiming at providing a flexible flow-based generative mechanism for synthesizing time-to-event data as well as learning their corresponding probabilistic densities. Unlike existing methods, CFM does not rely on survival likelihoods to handle censoring, but instead exploits the monotonicity property of differential flows to address censoring via pullback sampling. Under this framework, we implement an alternating procedure between pullback sampling and CFM steps, leading to a flow-based expectation--maximization (EM) algorithm tailored for time-to-event inference. Under mild distributional conditions, we establish the convergence of CFM to the true event-time distribution as the EM iterations proceed. We then apply CFM to generate time-to-event data and estimate their probabilistic densities, where the observed censoring time can be incorporated to provide truncated sampling or truncated density estimation for survival inference. Through extensive simulation studies, we demonstrate the benefits of CFM for complex and large-scale data generation. Finally, we apply CFM to analyze data from the Organ Procurement and Transplantation Network (OPTN) and the Medical Information Mart for Intensive Care IV (MIMIC-IV), showcasing its benefits for synthetic and predictive survival analyses in terms of data fidelity, predictive performance, and empirical privacy characteristics.
ER -