Research map / Optimization / Nonconvex and smooth optimization
Near-Optimal Nonconvex-Strongly-Convex Bilevel Optimization with Fully First-Order Oracles
Lesi Chen, Yaohua Ma, Jingzhao Zhang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use two-time-scale step sizes (larger step in x than y) in the penalty-based first-order method | F2BA vs F2SA · Deterministic first-order stationary point | oracle complexity: \epsilon^{-3} -> \epsilon^{-2} | Improved |
| use two-time-scale step sizes in the stochastic penalty method | F2BSA vs F2SA · Partially stochastic | oracle complexity: \epsilon^{-5} -> \epsilon^{-4} | Improved |
| use two-time-scale step sizes in the stochastic penalty method | F2BSA vs F2SA · Fully stochastic | oracle complexity: \epsilon^{-7} -> \epsilon^{-6} | Improved |
| add Nesterov acceleration with restart to perturbed F2BA | PF2BA · Second-order stationary point | oracle complexity: \epsilon^{-2} -> \epsilon^{-1.75} | Improved |
| use two-time-scale step sizes instead of single-time-scale for GPT-2 data hyper-cleaning | F2BSA vs F2SA and PBGD · Alpaca data hyper-cleaning, GPT-2 124M | convergence speed | Improved |
| use penalty-based method instead of HVP-based AID | F2BA · 20 newsgroup logistic regression, 100,000 regularizers | validation performance | Improved |
| add acceleration to F2BA | F2BA · 20 newsgroup logistic regression | validation performance | Improved |
Closest papers by meaning
- On The Complexity of First-Order Methods in Stochastic Bilevel Optimization (ICML 2024 · similarity 0.92)
- Penalty-Based First-Order Methods for Bilevel Optimization with Minimax and Constrained Lower-Level Problems (NeurIPS 2026 · similarity 0.90)
- On Penalty Methods for Nonconvex Bilevel Optimization and First-Order Stochastic Approximation (ICLR 2024 · similarity 0.90)
- An Accelerated Gradient Method for Convex Smooth Simple Bilevel Optimization (NeurIPS 2024 · similarity 0.89)
- Penalty-based Methods for Simple Bilevel Optimization under Hölderian Error Bounds (NeurIPS 2024 · similarity 0.89)
- First-Order Methods for Linearly Constrained Bilevel Optimization (NeurIPS 2024 · similarity 0.89)
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.
