A0600
Title: Incorporating gamma frailties into deep neural networks for clustered survival analysis
Authors: Hangbin Lee - Chungnam National University (Korea, South) [presenting]
Il Do Ha - Pukyong National University (Korea, South)
Youngjo Lee - Seoul National University (Korea, South)
Abstract: Prediction of time-to-event outcomes such as recurrence or survival is central to biomedical applications. Deep neural networks with gamma frailties are presented that address the challenges posed by clustered survival data. By introducing gamma frailties into neural networks and adopting a nonparametric baseline hazard, the proposed model effectively captures within-cluster dependencies. The method is based on the new h-likelihood formulation that unifies inference for both fixed and random effects, enabling scalable training. Through theoretical justification and comprehensive experiments, it is shown that the proposed model delivers robust cluster-specific predictions across a range of frailty distributions.
