AI Paper Atlas

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

RecurJac: An Efficient Recursive Algorithm for Bounding Jacobian Matrix of Neural Networks and Its Applications

Huan Zhang, Pengchuan Zhang, Cho-Jui Hsieh

AAAI 2019 · Adversarial robustness and security › Certified and empirical robustness · approach: Neural network robustness verification · 57 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 RecurJac instead of FastLipMNIST 3-layer network · runner-up target, undefendedcertified lower bound for ℓ∞ distortion: 0.01802 -> 0.02256Improved
Use RecurJac instead of FastLipMNIST 3-layer network · runner-up target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.09639 -> 0.11573Improved
Use RecurJac instead of FastLipMNIST 3-layer network · random target, undefendedcertified lower bound for ℓ∞ distortion: 0.02374 -> 0.02870Improved
Use RecurJac instead of FastLipMNIST 3-layer network · random target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.11753 -> 0.13753Improved
Use RecurJac instead of FastLipMNIST 3-layer network · least-likely target, undefendedcertified lower bound for ℓ∞ distortion: 0.02720 -> 0.03205Improved
Use RecurJac instead of FastLipMNIST 3-layer network · least-likely target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.14067 -> 0.16153Improved
Use RecurJac instead of FastLipMNIST 4-layer network · runner-up target, undefendedcertified lower bound for ℓ∞ distortion: 0.01602 -> 0.02104Improved
Use RecurJac instead of FastLipMNIST 4-layer network · runner-up target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.04232 -> 0.07350Improved
Use RecurJac instead of FastLipMNIST 4-layer network · random target, undefendedcertified lower bound for ℓ∞ distortion: 0.01882 -> 0.02399Improved
Use RecurJac instead of FastLipMNIST 4-layer network · random target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.05267 -> 0.08603Improved
Use RecurJac instead of FastLipMNIST 4-layer network · least-likely target, undefendedcertified lower bound for ℓ∞ distortion: 0.02018 -> 0.02519Improved
Use RecurJac instead of FastLipMNIST 4-layer network · least-likely target, adversarial trainingcertified lower bound for ℓ∞ distortion: 0.06417 -> 0.09863Improved

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