AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Adversarially Robust Change Point Detection

Mengchu Li, Yi Yu

NeurIPS 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Attack evaluation and defense critique · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use ARC instead of non-robust or isolated-outlier-robust detectors under strong attacks that create spurious change pointsARC / aARC · Simulation setting (i), spurious change points, n=5000|K_hat - K|Improved
Use ARC instead of competitors when adversary hides change points with large epsilon and small kappaARC / aARC · Simulation setting (ii), hiding change points|K_hat - K|Improved
Use robust ARC when competitors already detect change points correctlyARC · Simulation localisation errorHausdorff distance d_H / nHurt
Choose epsilon automatically via tournament procedure instead of using the true epsilon (ablation)ARC · Simulations|K_hat - K| and d_HNo clear effect
Compare ARC with Biweight and R_cusum on well-log dataaARC · Well-log data setdetected change pointsNo clear effect
Perturb Beijing PM2.5 data with injected outliers, compare aARC to Biweight(3) and R_cusumaARC · Beijing PM2.5 perturbeddetected change pointsImproved
Input epsilon = 0.01 instead of automatic epsilon (ablation)ARC · London PM2.5detected change pointsMixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy