AI Paper Atlas

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

On Collective Robustness of Bagging Against Data Poisoning

ICML 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness to poisoning and multiple attacks · 4 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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
Replace vanilla random subsampling with deterministic hash-based subsamplingBagging · Bank, 5% poison budget, G=20certified collective robustness: 4449 -> 9718Helped
Replace vanilla random subsampling with deterministic hash-based subsamplingBagging · Electricity, 5% poison budget, G=20certified collective robustness: 9348 -> 9915Helped
Replace vanilla random subsampling with deterministic hash-based subsamplingBagging · FMNIST and CIFAR-10, 5% poison budget, G=50certified collective robustness: FMNIST: 7727 -> 9726; CIFAR-10: 3621 -> 8393Helped
Use collective certification instead of sample-wise certificationHash bagging · Bank, 5% poison budget, G=20certified collective robustness: 9599 -> 9718Helped
Use collective certification with decomposition instead of collective certification alone (ablation)Hash bagging · FMNIST and CIFAR-10, 5% poison budget, G=50certified collective robustness: FMNIST: 9726 -> 9666; CIFAR-10: 8393 -> 8694Helped
Increase the number of sub-trainsets G toward GK=N (ablation)Hash bagging · Certified collective robustness across tested G valuescertified collective robustnessMixed
Increase sub-trainset size K toward K=N/G (ablation)Hash bagging · Certified collective robustness across tested K valuescertified collective robustnessMixed
Increase the sub-testset scale Δ (ablation)Hash bagging certification · Certified collective robustness without a time limitcertified collective robustnessHelped
Increase the sub-testset scale Δ under a 2s-per-sample time limit (ablation)Hash bagging certification · Certified collective robustness with limited solver timecertified collective robustnessMixed

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