AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Multi-block Min-max Bilevel Optimization with Applications in Multi-task Deep AUC Maximization

Quanqi Hu, YONGJIAN ZHONG, Tianbao Yang

NeurIPS 2022 · Optimization › Nonconvex and smooth optimization · approach: Penalty and implicit-gradient bilevel methods · 1 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 compositional training instead of directly optimizing the multi-task min-max AUC lossMulti-task deep AUC maximization · CIFAR100AUC: 0.9044 (0.0015) -> 0.9272 (0.0014)Improved
Use compositional training instead of directly optimizing the multi-task min-max AUC lossMulti-task deep AUC maximization · CheXpertAUC: 0.8084( 0.1455) -> 0.8198(0.1495)Improved
Use compositional training instead of directly optimizing the multi-task min-max AUC lossMulti-task deep AUC maximization · CelebAAUC: 0.9062 (0.0042) -> 0.9192 (0.0004)Improved
Use compositional training instead of directly optimizing the multi-task min-max AUC lossMulti-task deep AUC maximization · ogbg-molpcbaAUC: 0.7793(0.0028) -> 0.8406(0.0044)Improved
Use the proposed multi-task bilevel method instead of SOPA-sMulti-task deep partial AUC maximization · CIFAR100partial AUC (FPR≤0.1): 0.9251 (0.0003) -> 0.9262 (0.0005)Improved
Use the proposed multi-task bilevel method instead of SOPA-sMulti-task deep partial AUC maximization · CelebApartial AUC (FPR≤0.1): 0.8336 (0.0001) -> 0.8360 (0.0003)Improved
Use the proposed multi-task bilevel method instead of SOPA-sMulti-task deep partial AUC maximization · CheXpertpartial AUC (FPR≤0.1): 0.6682 (0.0156) -> 0.6827 (0.0183)Improved
Use the proposed multi-task bilevel method instead of SOPA-sMulti-task deep partial AUC maximization · ogbg-molpcbapartial AUC (FPR≤0.1): 0.7290 (0.0019) -> 0.7374 (0.0015)Improved
Increase the data-sample batch size (ablation)Proposed multi-task deep AUC and partial AUC methods · Multi-task AUC and partial AUC trainingconvergence speedImproved
Increase the task-sample batch size (ablation)Proposed multi-task deep AUC and partial AUC methods · Multi-task AUC and partial AUC trainingconvergence speedImproved

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