AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Joint Training of Deep Ensembles Fails Due to Learner Collusion

Alan Jeffares, Tennison Liu, Jonathan Crabbé, Mihaela van der Schaar

NeurIPS 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Active learning and subset selection · 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
train ensemble jointly on ensemble loss instead of independentlyResNet-18 ensemble · ImageNet valtop-1 accuracy: 68.10 -> 62.74Hurt
train ensemble jointly on ensemble loss instead of independentlyResNet-34 ensemble · ImageNet valtop-1 accuracy: 73.11 -> 68.91Hurt
train ensemble jointly on ensemble loss instead of independentlyDenseNet ensemble · ImageNet valtop-1 accuracy: 70.62 -> 60.81Hurt
train ensemble jointly on ensemble loss instead of independentlyShuffleNet V2 0.5x ensemble · ImageNet valtop-1 accuracy: 61.00 -> 53.41Hurt
train ensemble jointly on ensemble loss instead of independentlyShuffleNet V2 1x ensemble · ImageNet valtop-1 accuracy: 68.56 -> 61.96Hurt
train ensemble jointly on ensemble loss instead of independentlySqueezeNet V1.1 ensemble · ImageNet valtop-1 accuracy: 56.25 -> 42.69Hurt
train ensemble jointly on ensemble loss instead of independentlyMobileNet V3-small ensemble · ImageNet valtop-1 accuracy: 67.43 -> 57.34Hurt
train ensemble jointly on ensemble loss instead of independentlyMobileViT ensemble · ImageNet valtop-1 accuracy: 61.29 -> 54.69Hurt
increase beta toward 1 (more joint training) (ablation)ResNet ensembles · CIFAR-10/100test accuracyHurt
increase beta toward 1 (more joint training) (ablation)ResNet-18 base learners · ImageNetvalidation accuracyHurt

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