AI Paper Atlas

Research map / Optimization / Optimizers

A Gradient Method for Multilevel Optimization

Ryo Sato, Mirai Tanaka, Akiko Takeda

NeurIPS 2021 · Optimization › Optimizers · approach: Hypergradient bilevel optimization · 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
add attacker level (trilevel instead of bilevel hyperparameter model)multilevel gradient method · diabetes, early-stopped, sigma=0.08test MSE: 1.0573 -> 0.8601Improved
add attacker levelmultilevel gradient method · Boston, sigma=0.08test MSE: 0.4899 -> 0.4333Improved
add attacker levelmultilevel gradient method · wine (red), sigma=0.08test MSE: 0.7277 -> 0.7223Improved
add attacker levelmultilevel gradient method · wine (white), sigma=0.08test MSE: 0.8750 -> 0.8659Improved
trilevel instead of bilevel under growing test noisemultilevel gradient method · diabetes, Figure 3test MSE growth with noiseImproved
reduce lower-level iterations to T2=1, T3=1 from T2=10, T3=10 (ablation)Algorithm 1 · three-level toy problem (8)convergence to optimum (0,0)No clear effect
compare Algorithm 1 with evolutionary strategy method [23]Algorithm 1 · toy problem (8)solution qualityImproved
set lower-level iterations T2, T3 smaller on problems with complex inter-level relationshipsAlgorithm 1 · hyperparameter optimizationefficiencyMixed

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