Cascade-KDE: Robust Time-Series Restoration under Out-of-Distribution Impulse Corruptions

Yuefeng Liu, Ning Yang, Ziyu Yang

arXiv:2605.24055 · 2026-05-26 공개 · arXiv · PDF

training-free out-of-distribution battery-degradation density-truncated exponential-cascade ecg-analysis derivative-preservation bounded-density

Abstract

Real-world time-series data in industrial sensing, healthcare, and energy systems is often corrupted by a mixture of Gaussian noise and occasional large-magnitude impulse outliers. For tasks that depend on local shape, such as ECG morphology analysis and battery degradation monitoring, the main requirement is not only low reconstruction error but also preservation of derivative peaks and task-critical features. We propose Cascade-KDE, a training-free restoration framework for corrupted time series. The method first estimates a two-dimensional temporal-amplitude density, then applies a Density-Truncated Robust Expectation to limit the influence of distant abnormal points, and finally refines the sequence through an exponential cascade with adaptive stopping. This design aims to improve robustness under out-of-distribution impulse corruptions while keeping the restored trajectory close to the original local structure. Across several benchmark datasets, the proposed method shows consistent gains over classical filters and representative learning-based baselines on curve fidelity, derivative preservation, downstream classification, and runtime efficiency. These results suggest that bounded density-based restoration is a practical option for feature-preserving preprocessing in noisy time-series pipelines.

한국어 요약

📋 한 줄 요약

**[시계열 복원 / 충격 노이즈]** Cascade-KDE가 2D temporal-amplitude density + Density-Truncated Robust Expectation + adaptive cascade로 학습 없이 OOD 임펄스 노이즈 시계열 복원 — ECG·배터리 모니터링의 derivative peak 보존·고속.

🎯 핵심 기여도

💡 핵심 아이디어

Impulse-noise·OOD outlier가 섞인 시계열 복원에는 학습 기반 모델이 아니라 (1) temporal-amplitude density 추정 + (2) density-truncated robust expectation으로 outlier 영향 절단 + (3) adaptive stop을 가진 exponential cascade로 정제가 효과적이며, training-free이면서 derivative-preserving 복원이 가능하다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: OOD impulse-noise에 robust한 training-free 시계열 전처리, derivative·local shape 보존이라는 명확한 task 친화 목표, runtime·downstream 동시 우수로 실용 가치 큼. **한계**: KDE 기반으로 매우 긴 시계열·고차원 multivariate에서의 비용, adaptive stop·truncation hyperparameter 튜닝 필요, 비-impulse 노이즈에서의 ablation 추가 검증 여지.

🚀 실용적 활용