AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Beyond the Single Neuron Convex Barrier for Neural Network Certification

Gagandeep Singh, Rupanshu Ganvir, Markus Püschel, Martin Vechev

NeurIPS 2019 · Architectures and efficiency › RNNs and normalization · approach: Theory of depth and width · 78 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
relax multiple ReLUs jointly (k=2,3) instead of single-neuron triangle (ablation)kPoly · MNIST ConvSmall first 100 regionsverified regions (#): 20 -> 35Improved
increase k from 1 to 2 to 3 (ablation)kPoly · MNIST 9x200 output box volumeoutput box volume: 4.5272 · 1014 -> 2.9679 · 105Improved
group neurons by triangle area heuristic instead of random partition (ablation)kPoly · MNIST ConvSmall first 100 regionsverified regions (#): 34 -> 35Improved
add k-ReLU refinement to DeepPolyDeepPoly · MNIST 9x200verified regions (#): 259 -> 506Improved
add k-ReLU refinement to DeepPolyDeepPoly · MNIST ConvSmallverified regions (#): 158 -> 347Improved
add k-ReLU refinement to DeepPolyDeepPoly · CIFAR10 ResNetverified regions (#): 243 -> 245No clear effect
add k-ReLU refinement to DeepPolyDeepPoly · MNIST ConvBigverified regions (#): 711 -> 736Improved
kPoly vs RefineZonoverifier · MNIST ConvBigverified regions (#): 648 -> 736Improved
kPoly vs RefineZonoverifier · MNIST 9x200verified regions (#): 316 -> 506Improved

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