🤖 AI Summary
Medical survival analysis faces a fundamental trade-off between predictive performance and clinical interpretability: deep learning models achieve high accuracy but lack transparency, while traditional Cox models are interpretable yet limited in capturing complex hazard structures. To address this, we propose CoxKAN—the first end-to-end interpretable survival model integrating Kolmogorov–Arnold Networks (KANs) into the Cox proportional hazards framework. Our contributions are fourfold: (1) the first application of KANs to survival analysis; (2) automatic learning of intricate nonlinear covariate interactions, yielding human-readable symbolic expressions of the hazard function; (3) joint differentiable optimization of grid-based activation parameters and survival loss; and (4) end-to-end interpretable feature selection. Experiments demonstrate that CoxKAN exactly recovers ground-truth hazard formulas on four synthetic datasets, significantly outperforms the standard Cox model on nine real-world medical datasets, matches or exceeds tuned MLPs in predictive performance, and quantitatively reveals nonlinear risk effects of key biomarkers.
📝 Abstract
Survival analysis is a branch of statistics used for modeling the time until a specific event occurs and is widely used in medicine, engineering, finance, and many other fields. When choosing survival models, there is typically a trade-off between performance and interpretability, where the highest performance is achieved by black-box models based on deep learning. This is a major problem in fields such as medicine where practitioners are reluctant to blindly trust black-box models to make important patient decisions. Kolmogorov-Arnold Networks (KANs) were recently proposed as an interpretable and accurate alternative to multi-layer perceptrons (MLPs). We introduce CoxKAN, a Cox proportional hazards Kolmogorov-Arnold Network for interpretable, high-performance survival analysis. We evaluate the proposed CoxKAN on 4 synthetic datasets and 9 real medical datasets. The synthetic experiments demonstrate that CoxKAN accurately recovers interpretable symbolic formulae for the hazard function, and effectively performs automatic feature selection. Evaluation on the 9 real datasets show that CoxKAN consistently outperforms the Cox proportional hazards model and achieves performance that is superior or comparable to that of tuned MLPs. Furthermore, we find that CoxKAN identifies complex interactions between predictor variables that would be extremely difficult to recognise using existing survival methods, and automatically finds symbolic formulae which uncover the precise effect of important biomarkers on patient risk.