2021/03/08 by Jiaye Teng, Teng, Jiaye, Zeren Tan +3 · 3 citations
Computer Science · Mathematics · #Artificial Intelligence (cs.AI) #Data Structures and Algorithms (cs.DS) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning and Algorithms #Machine Learning and Data Classification #Statistical Methods and Bayesian Inference #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.2103.04556
openalex publication_date 2021/03/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
It is challenging to deal with censored data, where we only have access to\nthe incomplete information of survival time instead of its exact value.\nFortunately, under linear predictor assumption, people can obtain guaranteed\ncoverage for the confidence band of survival time using methods like Cox\nRegression. However, when relaxing the linear assumption with neural networks\n(e.g., Cox-MLP (Katzman et al., 2018; Kvamme et al., 2019)), we lose the\nguaranteed coverage. To recover the guaranteed coverage without linear\nassumption, we propose two algorithms based on conformal inference. In the\nfirst algorithm WCCI, we revisit weighted conformal inference and introduce a\nnew non-conformity score based on partial likelihood. We then propose a\ntwo-stage algorithm T-SCI, where we run WCCI in the first stage and apply\nquantile conformal inference to calibrate the results in the second stage.\nTheoretical analysis shows that T-SCI returns guaranteed coverage under milder\nassumptions than WCCI. We conduct extensive experiments on synthetic data and\nreal data using different methods, which validate our analysis.\n