Deep Lifetime Clustering
This work addresses the problem of clustering subjects based on lifetime distributions for researchers in survival analysis, offering a robust method that handles unobservable termination signals without assuming proportional hazards, though it is incremental in improving existing clustering approaches.
The paper tackles the problem of lifetime clustering by introducing a neural-network model that maximizes divergence between empirical lifetime distributions of clusters, using a novel loss based on a tight upper bound of the two-sample Kuiper test p-value. Results show significantly better performance on real and synthetic datasets, as measured by metrics like C-index and adjusted Rand index.
The goal of lifetime clustering is to develop an inductive model that maps subjects into $K$ clusters according to their underlying (unobserved) lifetime distribution. We introduce a neural-network based lifetime clustering model that can find cluster assignments by directly maximizing the divergence between the empirical lifetime distributions of the clusters. Accordingly, we define a novel clustering loss function over the lifetime distributions (of entire clusters) based on a tight upper bound of the two-sample Kuiper test p-value. The resultant model is robust to the modeling issues associated with the unobservability of termination signals, and does not assume proportional hazards. Our results in real and synthetic datasets show significantly better lifetime clusters (as evaluated by C-index, Brier Score, Logrank score and adjusted Rand index) as compared to competing approaches.