AI Paper Atlas

Research map / Optimization / Nonconvex and smooth optimization

Convergence Analysis of Decentralized Hessian-/Jacobian-Free Algorithm for Nonconvex Stochastic Bilevel Optimization

Yihan Zhang, Xinwen Zhang, My T. Thai, Jie Wu, Hongchang Gao

ICML 2026 · Optimization › Nonconvex and smooth optimization · approach: Penalty and implicit-gradient bilevel methods

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
increase the learning rate eta (ablation)proposed decentralized first-order bilevel algorithm · hyperparameter optimization on a9aupper-level loss convergenceMixed
vary the momentum coefficient (ablation)proposed decentralized first-order bilevel algorithm · hyperparameter optimization on a9aupper-level loss convergenceNo clear effect
use a smaller penalty parameter rho (larger penalty 1/rho) (ablation)proposed decentralized first-order bilevel algorithm · hyperparameter optimization on a9aupper-level loss convergenceImproved
use the proposed algorithm instead of F2BSA variants (F2BSA-1, -5, -10, -VR)decentralized bilevel optimization · hyperparameter optimization (a9a, w8a, covtype, MNIST)upper-level loss vs iterationsImproved
use the proposed algorithm instead of F2BSA variantsdecentralized bilevel optimization · model pruning ResNet20 CIFAR10 (20% parameters kept)upper-level loss vs iterationsImproved
use the fully first-order algorithm instead of second-order DSBO, gossip-DSBO, MA-DSBO, VRDBO, DSVRBGD-Sstrongly convex lower-level hyperparameter optimization · logistic regression hyperparameter optimizationtime to convergenceImproved
add variance-reduced gradients relative to DSFDA-GTfully first-order decentralized bilevel method · logistic regression hyperparameter optimizationconvergenceImproved

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