Accurate Evaluation of Quickest Changepoint Detectors via Non-parametric Survival Analysis
Summary
The paper proposes non-parametric estimators KM-ARL and KM-ADD for evaluating changepoint detectors under finite and irregular sequence lengths, drawing an analogy between QCD and survival analysis.
View Cached Full Text
Cached at: 05/20/26, 08:35 AM
# Accurate Evaluation of Quickest Changepoint Detectors via Non-parametric Survival Analysis Source: [https://arxiv.org/abs/2605.18798](https://arxiv.org/abs/2605.18798) [View PDF](https://arxiv.org/pdf/2605.18798) > Abstract:We propose non\-parametric estimators for the average run length \(ARL\) and average detection delay \(ADD\) in quickest changepoint detection \(QCD\) under finite and irregular sequence lengths\. Although ARL and ADD are widely used as optimality criteria in theoretical and simulation studies, their application to real\-world datasets is hindered by limited and irregular sequence lengths\. To address this issue, we propose non\-parametric estimators for the ARL and ADD, termed KM\-ARL and KM\-ADD, by drawing an analogy between QCD and survival analysis to model detection probabilities under sequence truncation\. We derive estimation bias bounds and prove that they are asymptotically unbiased unless extrapolation is required\. Experiments on simulated and real\-world datasets demonstrate their practical utility, enhancing robustness against limited and irregular sequence lengths, improving interpretability, and facilitating empirical, intuitive model selection\. Our Python code is provided at[this https URL](https://github.com/TaikiMiyagawa/Kaplan-Meier-Average-Run-Length), offering ready\-to\-use implementations for practitioners\. ## Submission history From: Taiki Miyagawa \[[view email](https://arxiv.org/show-email/ca4d58af/2605.18798)\] **\[v1\]**Mon, 11 May 2026 07:23:48 UTC \(1,659 KB\)
Similar Articles
Aligning Data-Driven Predictors with Allocation: A Decision-Focused Approach to Survival Analysis
This paper introduces a decision-focused learning approach for survival analysis that aligns predictive models with downstream allocation decisions, using NDCG optimization. Applied to US heart transplant data, it improves ranking performance by 50-100%, potentially yielding thousands of additional life-years annually.
Calibration, Uncertainty Communication, and Deployment Readiness in CKD Risk Prediction: A Framework Evaluation Study
This study evaluates five machine learning classifiers for chronic kidney disease risk prediction, finding that near-perfect internal performance fails under distribution shift. It emphasizes the need for calibration stability and conformal coverage transfer before clinical deployment.
CALAD: Channel-Aware contrastive Learning for multivariate time series Anomaly Detection
Proposes CALAD, a channel-aware contrastive learning framework for multivariate time series anomaly detection that uses estimated channel relevance to construct contrastive samples, achieving state-of-the-art performance.
Robust Checkpoint Selection for Multimodal LLMs via Agentic Evaluation and Stability-Aware Ranking
This paper addresses the challenge of robust checkpoint selection for multimodal LLMs under evaluation uncertainty, proposing a multi-stage framework that integrates curated real-world data, LLM-based judgment, and ranking protocols with confidence estimation.
Cascade-KDE: Robust Time-Series Restoration under Out-of-Distribution Impulse Corruptions
Proposes Cascade-KDE, a training-free framework that uses two-dimensional kernel density estimation and truncated expectation to restore time-series corrupted by out-of-distribution impulse outliers while preserving local structure and derivative features.