AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Loss Surface Simplexes for Mode Connecting Volumes and Fast Ensembling

Gregory Benton, Wesley Maddox, Sanae Lotfi, Andrew Gordon Gordon Wilson

ICML 2021 · Architectures and efficiency › RNNs and normalization · approach: Symmetry and equivariance in loss landscapes · 4 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
Add a 2-simplex at each of three independently trained ensemble componentsVGG-16 deep ensemble · CIFAR-10test error: 6.2% -> 5.7%Improved
Use ESPRO simplexes instead of independent models at a fixed training budgetVGG-16 ensemble · CIFAR-10 and CIFAR-100test errorImproved
Use ESPRO rather than a standard deep ensemble at a fixed ensemble sizeVGG-16 ensemble · CIFAR-10 and CIFAR-100test errorImproved
Increase simplex order within ESPRO (ablation)VGG-16 ensemble · CIFAR-10 and CIFAR-100test errorImproved
Increase the number of independent ensemble components (ablation)ESPRO and deep ensembles · CIFAR-10test errorImproved
Use a SPRO simplex rather than a deep ensemble for transformer modelsImage transformer · CIFAR-100test errorNo clear effect
Use a four-dimensional ESPRO simplex instead of a deep ensemble (ablation)Image transformer · CIFAR-100test errorImproved
Use a four-dimensional ESPRO simplex instead of a deep ensemble (ablation)Image transformer · CIFAR-100NLLHurt
Ensemble over a mode-connecting simplex complex rather than individual models (ablation)SPRO mode-connecting complex · CIFAR-10 and CIFAR-100test errorNo clear effect
Add a connector after the first ten connectors (ablation)SPRO mode-connecting complex · CIFAR-10, VGG-16simplex volume: 10^5 -> approximately 10^{-4}Hurt
Increase simplex sample count beyond approximately 25 (ablation)SPRO ensemble · CIFAR-100test errorNo clear effect
Apply temperature scaling to predictions under Gaussian noise corruptionESPRO and comparison methods · CIFAR-10 Gaussian noise corruptionNLLImproved

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