AI Paper Atlas

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

Training Over-parameterized Models with Non-decomposable Objectives

Harikrishna Narasimhan, Aditya Krishna Menon

NeurIPS 2021 · Learning theory and algorithms › Prediction and decision losses · approach: Losses for decision-making · 0 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
Replace loss re-weighting with logit-adjusted loss for the diagonal-gain cost-sensitive stepResNet cost-sensitive learning · CIFAR-10-LTMin Rec: 0.592 -> 0.721Improved
Replace loss re-weighting with logit-adjusted loss for the diagonal-gain cost-sensitive stepResNet cost-sensitive learning · CIFAR-100-LTMin HT Rec: 0.083 -> 0.418Improved
Replace loss re-weighting with logit-adjusted loss for the diagonal-gain cost-sensitive stepResNet cost-sensitive learning · TinyImgNet-LTMin HT Rec: 0.098 -> 0.310Improved
Use hybrid cost-sensitive loss instead of re-weighted lossResNet cost-sensitive learning · CIFAR-10-LTAvg Rec: 0.739 -> 0.759Improved
Use hybrid cost-sensitive loss instead of re-weighted lossResNet cost-sensitive learning · CIFAR-100-LTAvg Rec: 0.408 -> 0.474Improved
Use hybrid cost-sensitive loss instead of re-weighted lossResNet cost-sensitive learning · TinyImgNet-LTAvg Rec: 0.289 -> 0.350Improved
Use hybrid loss with D = diag(1/π) rather than D = diag(Gii) (ablation)CSL hybrid loss · CIFAR-10-LTAvg Rec: 0.759 vs 0.760No clear effect
Use hybrid loss with self-distillation rather than ordinary training (ablation)CSL logit-adjusted method · CIFAR-10-LTMin Rec: 0.721 -> 0.734Improved
Use hybrid loss with self-distillation rather than ordinary training (ablation)CSL hybrid method · CIFAR-100-LTMin HT Rec: 0.418 -> 0.467Improved
Train with proposed cost-sensitive loss instead of ERM, then apply post-shiftingCIFAR-100-LT worst-case recall over all classes · CIFAR-100-LTMin Rec: 0.090 -> 0.120Mixed

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