AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs

Timur Garipov, Pavel Izmailov, Dmitrii Podoprikhin, Dmitry P Vetrov, Andrew G Wilson

NeurIPS 2018 · Architectures and efficiency › RNNs and normalization · approach: Symmetry and equivariance in loss landscapes · 203 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 FGE instead of Snapshot EnsemblesVGG-16 · CIFAR-100, 1B budgeterror rate (%): 26.4 ± 0.1 -> 25.7 ± 0.1Improved
Use FGE instead of Snapshot EnsemblesResNet-164 · CIFAR-100, 1B budgeterror rate (%): 20.9 ± 0.2 -> 20.2 ± 0.1Improved
Use FGE instead of Snapshot EnsemblesWRN-28-10 · CIFAR-100, 1B budgeterror rate (%): 17.9 ± 0.2 -> 17.7 ± 0.2Improved
Use FGE instead of Snapshot EnsemblesVGG-16 · CIFAR-10, 1B budgeterror rate (%): 6.57 ± 0.12 -> 6.48 ± 0.09Improved
Use FGE instead of Snapshot EnsemblesResNet-164 · CIFAR-10, 1B budgeterror rate (%): 4.66 ± 0.02 -> 4.54 ± 0.05Improved
Use FGE instead of Snapshot EnsemblesWRN-28-10 · CIFAR-10, 1B budgeterror rate (%): 3.73 ± 0.04 -> 3.65 ± 0.1Improved
Fine-tune a pretrained model with FGE for 5 epochsResNet-50 · ImageNet ILSVRC-2012top-1 test error rate: 23.87 -> 23.31Improved
Replace the direct line segment between trained modes with a learned polygonal chain (ablation)Mode-connecting ResNet-164 · CIFAR-100test error (%): 53.69 -> 24.92Improved
Use a learned Bezier curve instead of a direct segment (ablation)Mode-connecting ResNet-164 · CIFAR-100test error (%): 53.69 -> 24.99Improved
Temperature-scale predictions from curve points (ablation)50-point curve ensemble · CIFAR-100test error rate: 21.03% -> 20.7%Improved
Ensemble 50 points along the learned curve rather than only its endpoints (ablation)ResNet-164 · CIFAR-100test error rate: 22.0% -> 21.03%Improved
Use a curve ensemble rather than an independent three-network ensembleResNet-164 · CIFAR-100test error rate: 21.03% vs 21.01%No clear effect

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