K-Survival Means

2026-07-27

AI Summary

This work addresses the limitation of existing survival data clustering methods, which often neglect survival outcome information and consequently fail to yield clusters with statistically significant survival differences. To overcome this, we propose K-SurvMeans, a novel approach that uniquely integrates survival outcomes directly into the K-Means objective function, optimizing it to maximize inter-cluster survival disparity. The method leverages deep dimensionality reduction to construct low-dimensional latent representations and employs particle swarm optimization to handle the non-differentiability of the objective function. Extensive experiments on multiple public survival datasets demonstrate that K-SurvMeans substantially outperforms current deep survival clustering techniques, achieving superior separation of survival distributions across clusters while maintaining computational efficiency and scalability.

Abstract

In this work, we propose K-SurvMeans, a novel extension of K-Means for clustering survival data. The method explicitly uses the survival outcome in the clustering process to optimize cluster centers, thereby maximizing pairwise survival differences between clusters. The objective function encourages the clusters to be well-separated from the survival perspective. Since the resulting optimization problem is non-differentiable, we employ the Particle Swarm algorithm for the Optimization process. To further improve flexibility and mitigate the curse of dimensionality, we extend the framework to operate in a learned low-dimensional latent space obtained via a dimensionality reduction. This allows the method to capture better-separated clusters and enhance optimization efficiency by reducing the search space. Experiments on multiple publicly available benchmark survival datasets demonstrate that K-SurvMeans consistently yields clusters with improved separation in survival distributions compared to existing deep learning-based survival clustering methods.

Problem

• survival clustering
• survival data
• cluster separation
• survival analysis
• K-Means extension

Innovation

• K-SurvMeans
• survival clustering
• Particle Swarm Optimization
• latent space
• dimensionality reduction