Research map / Optimization / Nonconvex and smooth optimization
On Penalty-based Bilevel Gradient Descent Method
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the function-value-gap penalty (V-PBGD) instead of gradient-norm penalty (G-PBGD) (ablation) | penalty-based bilevel GD · Data hyper-cleaning MNIST (2-layer MLP) | Test accuracy: 94.58 vs 92.17 | Helped |
| Use the function-value-gap penalty (V-PBGD) instead of gradient-norm penalty (G-PBGD) (ablation) | penalty-based bilevel GD · Data hyper-cleaning MNIST (linear model) | Test accuracy: 90.48 vs 90.09 | Helped |
| Penalty-based bilevel GD (V-PBGD) vs. baselines for data hyper-cleaning | V-PBGD · Data hyper-cleaning MNIST (2-layer MLP) | Test accuracy: 94.58 vs 91.72 (best baseline IAPTT-GM) | Helped |
| Penalty-based bilevel GD (V-PBGD) vs. IAPTT-GM | V-PBGD · Data hyper-cleaning MNIST (linear model) | Test accuracy: 90.48 vs 90.44 | No clear effect |
| Penalty-based bilevel GD (V-PBGD) vs. IAPTT-GM runtime | V-PBGD · Data hyper-cleaning MNIST (2-layer MLP) | Runtime (second): 207.53 vs 1310.63 s | Helped |
| Penalty-based bilevel GD (V-PBGD) vs. RHG memory | V-PBGD · Data hyper-cleaning MNIST (2-layer MLP) | GPU memory (MB): 1199 vs 7997 MB | Helped |
| Iteration count vs penalty constant gamma in V-PBGD on toy bilevel problem (ablation) | V-PBGD · Toy non-convex bilevel problem (12) | Iteration complexity: Theta(gamma) scaling | No clear effect |
Closest papers by meaning
- Penalty-based Methods for Simple Bilevel Optimization under Hölderian Error Bounds (NeurIPS 2024 · similarity 0.90)
- On Penalty Methods for Nonconvex Bilevel Optimization and First-Order Stochastic Approximation (ICLR 2024 · similarity 0.89)
- An Accelerated Gradient Method for Convex Smooth Simple Bilevel Optimization (NeurIPS 2024 · similarity 0.89)
- Penalty-Based First-Order Methods for Bilevel Optimization with Minimax and Constrained Lower-Level Problems (NeurIPS 2026 · similarity 0.89)
- Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization (ICML 2024 · similarity 0.89)
- Generalized Smooth Bilevel Optimization with Nonconvex Lower-Level (ICML 2025 · similarity 0.88)
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.