AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Training independent subnetworks for robust prediction

Marton Havasi, Rodolphe Jenatton, Stanislav Fort, Jeremiah Zhe Liu, Jasper Snoek, Balaji Lakshminarayanan, Andrew Mingbo Dai, Dustin Tran

ICLR 2021 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 28 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 MIMO with three subnetworks instead of a deterministic networkResNet28-10 · CIFAR10NLL: 0.159 -> 0.123107Improved
Use MIMO with three subnetworks instead of a deterministic networkResNet28-10 · CIFAR100accuracy: 79.8 -> 81.9819Improved
Use MIMO with two subnetworks and input repetition (ρ=0.6) instead of a deterministic networkResNet50 · ImageNetaccuracy: 76.1 -> 77.5Improved
Use MIMO with two subnetworks and input repetition (ρ=0.6) instead of a deterministic networkResNet50 · ImageNet-CECE: 0.105 -> 0.106Hurt
Increase the number of subnetworks; the ensemble improves before capacity limits individual members (ablation)MIMO ResNet28-10 · CIFAR10 and CIFAR100ensemble performanceMixed
Add correlated input repetition during training (ρ>0) (ablation)MIMO ResNet50 · ImageNetperformanceImproved
Increase input repetition toward ρ=1 (ablation)MIMO ResNet50 · ImageNetensemble benefitHurt
Increase minibatch repetition (ablation)MIMO · CIFAR and ImageNet experimentsperformanceImproved
Increase regularization strength (ablation)MIMO ResNet28-10 · CIFAR10 and CIFAR100accuracy and log-likelihoodHurt
Increase ensemble size under stronger regularization (ablation)MIMO ResNet28-10 · CIFAR10 and CIFAR100accuracy and log-likelihoodHurt
Use separate inputs for each output rather than shared input in a naive multihead network (ablation)MIMO · CIFAR10subnetwork diversityImproved

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