Research map / Optimization / Nonconvex and smooth optimization
On lp-hyperparameter Learning via Bilevel Nonsmooth Optimization
Takayuki Okuno, Akiko Takeda, Akihiro Kawana, Motokazu Watanabe
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 Algorithm 1 instead of bayesopt | bilevel l0.8 hyperparameter learning · Facebook, p=1 | time (sec): 17.399 vs 146.144 | Improved |
| use Algorithm 1 instead of bayesopt | bilevel l0.8 hyperparameter learning · Student, p=0.8 | time (sec): 2.348 vs 409.382 | Improved |
| use Algorithm 1 instead of gridsearch | bilevel l0.8 hyperparameter learning · BodyFat, p=0.8 | sparsity: 0.000 vs 0.429 | Hurt |
| use Algorithm 1 over bayesopt and gridsearch for validation error | bilevel lp hyperparameter learning · 15 problem instances | time (sec): best time in 10 out of 15 | Improved |
| decrease p from 0.8 or 1 to 0.5 | Algorithm 1 · Facebook | sparsity: 0.925 vs 0.057 and 0.245 | Improved |
| increase sample size of Facebook from 0.5 to full | bayesopt · Facebook, l0.8 | time (sec) | Hurt |
| increase feature size of Student | bayesopt and gridsearch · Student, l0.8 | time (sec) | Hurt |
| stop Algorithm 1 early when smoothing parameter mu is about 0.01 (ablation) | Algorithm 1 · Facebook and BodyFat, l0.8 | test error | Improved |
| use fmincon (Alg.1-B) instead of implicit function approach (Alg.1-A) for subproblem (ablation) | Algorithm 1 · Insurance, multiple hyperparameters | sparsity: 0.000 -> 0.988 | Improved |
| use fmincon (Alg.1-B) instead of Alg.1-A (ablation) | Algorithm 1 · Facebook, multiple hyperparameters | feasibility | Hurt |
| use Algorithm 1 instead of bayesopt with multiple hyperparameters | bilevel l0.5 problem · Student, 273 hyperparameters | Errval: 0.771 vs 18.324 | Improved |
| use Algorithm 1 instead of bayesopt with multiple hyperparameters | bilevel l0.5 problem · Facebook, 54 hyperparameters | time (sec): 20.247 vs 600.000 | Improved |
Closest papers by meaning
- TSP: A Two-Sided Smoothed Primal-Dual Method for Nonconvex Bilevel Optimization (ICML 2025 · similarity 0.89)
- Bilevel Optimization under Unbounded Smoothness: A New Algorithm and Convergence Analysis (ICLR 2024 · similarity 0.89)
- Generalized Smooth Bilevel Optimization with Nonconvex Lower-Level (ICML 2025 · similarity 0.88)
- Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization (ICML 2024 · similarity 0.88)
- Tuning-Free Bilevel Optimization: New Algorithms and Convergence Analysis (ICLR 2025 · similarity 0.88)
- Efficient Curvature-Aware Hypergradient Approximation for Bilevel Optimization (ICML 2025 · similarity 0.87)
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.
