A1178
Title: Deep generative estimation of conditional survival function
Authors: Xingyu Zhou - Dow Inc (United States)
Wen Su - The University of Hong Kong (Hong Kong)
Changyu Liu - The Hong Kong Polytechnic University (Hong Kong)
Yuling Jiao - Wuhan University (China)
Xingqiu Zhao - The Hong Kong Polytechnic University (Hong Kong) [presenting]
Jian Huang - The Hong Kong Polytechnic University (China)
Abstract: A deep generative approach is proposed to the nonparametric estimation of conditional survival and hazard functions with censored data. The key idea of the proposed method is first to learn a conditional generator for the joint conditional distribution of the observed time and censoring indicator given covariates and then construct the Kaplan-Meier and Nelson-Aalen estimators based on this conditional generator for conditional hazard and survival functions. The method combines ideas from the recently developed deep generative learning and classical nonparametric estimation in survival analysis. The convergence properties of the generative nonparametric estimators are established. The numerical studies validate the proposed method.
