Research map / Learning theory and algorithms / Prediction and decision losses
A nonparametric method for gradual change problems with statistical guarantees
Lizhen Nie, Dan L Nicolae
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use max-gap estimator ρ̌ instead of arg-max estimator ρ̂ | proposed method · Setting I quadratic location | average l1 localization error: 0.24 -> 0.15 | Improved |
| use max-gap estimator ρ̌ instead of arg-max estimator ρ̂ | proposed method · Setting I volatility linear | average l1 localization error: 0.26 -> 0.15 | Improved |
| use max-gap estimator ρ̌ instead of arg-max estimator ρ̂ | proposed method · Setting I location linear d=50 | average l1 localization error: 0.09 -> 0.05 | Improved |
| use max-gap estimator ρ̌ instead of arg-max estimator ρ̂ | proposed method · Setting II network linear | average l1 localization error: 0.07 -> 0.02 | Improved |
| use max-gap estimator ρ̌ instead of arg-max estimator ρ̂ | proposed method · Setting II location linear d=1 | average l1 localization error: 0.12 -> 0.14 | No clear effect |
| replace general-purpose ρ̂_gen with proposed ρ̌ | baseline ρ̂_gen · Setting I location linear d=1 | average l1 localization error: 0.17 -> 0.09 | Improved |
| replace abrupt-change KCpA with proposed ρ̌ | KCpA · Setting I location complex d=1 | average l1 localization error: 0.27 -> 0.03 | Improved |
| replace abrupt-change KCpA with proposed ρ̌ | KCpA · Setting II location complex d=1 | average l1 localization error: 0.29 -> 0.04 | Improved |
| replace abrupt-change KCpA with proposed ρ̌ | KCpA · Setting I network linear | average l1 localization error: 0.16 -> 0.10 | Improved |
| use assumption-specific ρ̂_one-side instead of ρ̌ when the change is not one-sided | ρ̂_one-side · Setting I location complex d=1 | average l1 localization error: 0.03 -> 0.62 | Hurt |
| use ρ̂_mix instead of ρ̌ as dimension grows | ρ̂_mix · Setting I location linear d=50 | average l1 localization error: 0.05 -> 0.43 | Hurt |
| use one-sided estimator ρ̂_one-side when its assumption holds | ρ̂_one-side · Setting I location one-sided d=1 | average l1 localization error: 0.03 -> 0.02 | Mixed |
Closest papers by meaning
- Detection and Localization of Changes in Conditional Distributions (NeurIPS 2022 · similarity 0.87)
- Reducing sequential change detection to sequential estimation (ICML 2024 · similarity 0.86)
- Divide and Conquer Dynamic Programming: An Almost Linear Time Change Point Detection Methodology in High Dimensions (ICML 2023 · similarity 0.86)
- Change point detection and inference in multivariate non-parametric models under mixing conditions (NeurIPS 2023 · similarity 0.86)
- Optimal Online Change Detection via Random Fourier Features (NeurIPS 2025 · similarity 0.85)
- The Sample Complexity of Multiple Change Point Identification under Bandit Feedback (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
